Pancreatic cancer was not included, \* intraoperative massive bleeding in 3 cases, PDJV; proximal-

Pancreatic cancer usually is found at an advanced stage of cancer. In spite of resectable pancreatic cancer, due to anatomic intimacy between the pancreas and surrounding major venous vascular structures, unexpected portal vein (PV) and superior mesenteric vein invasion (SMV) is often encountered during the operation. In the past, several articles [53,54] concluded that the oncologic outcome of PD with combined venous vascular resection (PD-VR) is not favorable, however some others showed that PD-VR did not increase the complication rate and postoperative

dorsal jejunal vein, TMpE; total mesopancreas excision, SMA; superior mesenteric artery

process from SMA lateral border 10 432 166 100 16.7 2 0

TMpE 59 NA NA 86 NA NA \* NA

Rho, 2018 [48] ICG-guided differentiation of uncinate

Morales, 2019 [50] Periadventitial dissection of for SMA

2.1.3. Combined Vascular Resection

**mL R0, % LOH,** 

**Day Cx, % Mx.,** 

**%** 

**Table 3.** Several approaches to obtain clear SMA lateral margin*.*

#### 2.1.3. Combined Vascular Resection

Pancreatic cancer usually is found at an advanced stage of cancer. In spite of resectable pancreatic cancer, due to anatomic intimacy between the pancreas and surrounding major venous vascular structures, unexpected portal vein (PV) and superior mesenteric vein invasion (SMV) is often encountered during the operation. In the past, several articles [53,54] concluded that the oncologic outcome of PD with combined venous vascular resection (PD-VR) is not favorable, however some others showed that PD-VR did not increase the complication rate and postoperative mortality [55,56]. PD-VR is no longer considered an absolute contraindication in pancreatic head cancer. It is regarded as a safe and feasible option to improve the resection rate of pancreatic cancer. Especially, the oncologic role of combined venous vascular resection became highlighted as the concept of neoadjuvant chemotherapy was introduced in managing borderline or locally advanced pancreatic cancer. However, Peng et al. [57] recently evaluated the value of PD with combine vascular resection (PD-VR) for pancreatic head cancer. In this meta-analysis, a total of 12,031 patients (2186 patients with PD-VR, 22.2%) from 30 published articles were investigated. In comparison with the PD group, it was analyzed that the PD-VR group had a lower R0 resection rate and higher rates of complications such as biliary fistula, reoperation rate, delayed gastric emptying, cardiopulmonary abnormalities, hemorrhage, in-hospital mortality, and 30-day mortality. In addition, the blood loss, operation time, and total length of hospital stay were higher in the PD-VR group, concluding that PD-VR for pancreatic cancer should be carefully considered by selected pancreatic surgeons.

Until now, only a few case reports [58–62] and case series [19,63–65] have been reported on the technical feasibility and safety of LPD with combined venous vascular resection (LPD-VR, Table 4). Tangential resection with stapler or hand suture or patch reconstruction appears to be common among the patients with LPD-VR (Figure 3), which is also confirmed when comparing with OPD-VR [19]. As shown above, postoperative morbidity and mortality related to combined vascular resection are not ignorable, and segmental resection and end-to-end anastomosis or artificial graft, or renal vein graft must be a challenging procedure. Therefore, LPD-VR also needs to be performed in well-selected patients and by highly selected expert surgeons.


**Table 4.** Summary of articles reporting more than 5 cases of LPD-venous vascular resection (LPD-VR) for pancreatic cancer.

T, tangential resection; E-E, end-to-end anastomosis; RVG, renal vein graft; AG, artificial graft.

mortality [55,56]. PD-VR is no longer considered an absolute contraindication in pancreatic head cancer. It is regarded as a safe and feasible option to improve the resection rate of pancreatic cancer. Especially, the oncologic role of combined venous vascular resection became highlighted as the concept of neoadjuvant chemotherapy was introduced in managing borderline or locally advanced pancreatic cancer. However, Peng et al. [57] recently evaluated the value of PD with combine vascular resection (PD-VR) for pancreatic head cancer. In this meta-analysis, a total of 12,031 patients (2186 patients with PD-VR, 22.2%) from 30 published articles were investigated. In comparison with the PD group, it was analyzed that the PD-VR group had a lower R0 resection rate and higher rates of complications such as biliary fistula, reoperation rate, delayed gastric emptying, cardiopulmonary abnormalities, hemorrhage, in-hospital mortality, and 30-day mortality. In addition, the blood loss, operation time, and total length of hospital stay were higher in the PD-VR group, concluding that

PD-VR for pancreatic cancer should be carefully considered by selected pancreatic surgeons.

selected patients and by highly selected expert surgeons.

Until now, only a few case reports [58–62] and case series [19,63–65] have been reported on the technical feasibility and safety of LPD with combined venous vascular resection (LPD-VR, Table 4). Tangential resection with stapler or hand suture or patch reconstruction appears to be common among the patients with LPD-VR (Figure 3), which is also confirmed when comparing with OPD-VR [19]. As shown above, postoperative morbidity and mortality related to combined vascular resection are not ignorable, and segmental resection and end-to-end anastomosis or artificial graft, or renal vein graft must be a challenging procedure. Therefore, LPD-VR also needs to be performed in well-

The long-term oncologic efficacy of LPD-VR is lacking, but Croome et al. [19] demonstrated no statistical significance between LPD-VR and OPD-VR in terms of overall survival, leaving some room to be further investigated. In addition, there are other modified techniques for reconstruction

[68], suggesting further potential applications of laparoscopic technique remain to be studied.

**Figure 3.** Combined venous vascular resection during laparoscopic pancreaticoduodenectomy (LPD). Tangential wedge resection of SMV is performed after transient clamping venous system (**a**). Primary repair of resected venous system (**b**). Tangential resection line (white dotted line), tumor invasion (white \*), primary repair (white arrows). **Figure 3.** Combined venous vascular resection during laparoscopic pancreaticoduodenectomy (LPD). Tangential wedge resection of SMV is performed after transient clamping venous system (**a**). Primary repair of resected venous system (**b**). Tangential resection line (white dotted line), tumor invasion (white \*), primary repair (white arrows).

The long-term oncologic efficacy of LPD-VR is lacking, but Croome et al. [19] demonstrated no statistical significance between LPD-VR and OPD-VR in terms of overall survival, leaving some room to be further investigated. In addition, there are other modified techniques for reconstruction following LPD-VR, using falciform ligament [66], parietal peritoneum [67], and hepatic ligament teres [68], suggesting further potential applications of laparoscopic technique remain to be studied.

#### 2.1.4. Repeated Pancreatectomy for Pancreatic Cancer in Remnant Pancreas

Even after R0 radical resection, patients with PC often experience local recurrence. However, it has been reported that about 0.5–4.6% of pancreatic cancers develop in remnant pancreas [69,70]. Since 1995, re-resection of pancreatic cancer in remnant pancreas has been proposed for improving oncologic outcome [71]. The first case series reporting 30 patients with surgery for recurred pancreatic cancer also suggested that resection for recurrent pancreatic cancer can be performed safely, and questioned a potential subgroup who might actually benefit form re-resection of recurred pancreatic cancer [72].

As experience of resected remnant pancreatic cancer accumulated, Yamada et al. [73] published the results of resection of recurrent remnant pancreatic cancer in Japanese society of hepato-biliary-pancreatic surgery. Clinical data from 114 patients with remnant pancreatic cancer after initial pancreatectomy were analyzed. It was found that median survival of the resected remnant pancreatic cancer was superior to the non-resected group (26 and 14 months, respectively [hazard ratio: 0.56, *p* = 0.012]), showing that re-resection of remnant pancreatic cancer could offer a favorable outcome and chance for cure. Similarly, the recent literature [69] reviewing 49 reported patients with resected remnant pancreatic cancer following resection of the primary pancreatic cancer demonstrated that median disease-free survival was 44.4 months (12–143 months), and median survival time was 32 months after repeated pancreatectomy. In addition, another systemic review demonstrated that re-resection of isolated local recurrent pancreatic cancer showed the most favorable survival outcomes (median, 32 months) comparing with chemoradiation therapy (19 months), stereotactic body radiation therapy (16 months) [74], suggesting the role of aggressive surgical extirpation of recurred remnant pancreatic cancer for improving prognosis. Table 5 summarized recent review articles reporting oncologic benefit of repeated pancreatectomy for pancreatic cancer. It was found that completion total pancreatectomy is the most common procedure for treating recurred pancreatic cancer in remnant pancreas and nearly no postoperative mortality is reported. Reported median survival time (14–32 months), and 5-year overall survival (40.6%) following repeated pancreatectomy are much longer than unresectable pancreatic cancer in recent RCTs [75].

Then, is laparoscopic repeated pancreatectomy feasible and does it provide a potential oncologic role in treating recurrent pancreatic cancer in remnant pancreas? Until now, very few reports have been published showing the technical feasibility of laparoscopic completion of total pancreatectomy for remnant recurrent pancreatic cancer following LPD for primary pancreatic cancer. Recently, Kang et al. (accepted in Ann Hepatobiliary Pancreat Surg 2020, and in process) successfully demonstrated the technical feasibility of laparoscopic repeated pancreatectomy for recurrent pancreatic cancer following initial laparoscopic radical pancreatectomy. In addition, Nagakawa et al. [76] and Sunagawa et al. [77] also showed the technical feasibility of laparoscopic resection of remnant pancreatic cancer after LPD for other primary cancer. All these reports are suggesting the technical feasibility and potential role of a minimally invasive approach even in recurrent pancreatic cancer in remnant pancreas. Further study is mandatory.

**Table 5.** Summary of review articles of repeated pancreatectomy for recurred pancreatic cancer in remnant pancreas.


Px, pancreatectomy; DP, distal pancreatectomy; NA, not available.

#### **3. Current Literature**

Recently, three meta-analyses have been published, comparing the oncologic outcomes of LPD and OPD. Jiang et al. [81] systematically reviewed the articles comparing LPD and OPD for the treatment of pancreatic cancer. They found 8 studies involving 15,278 patients and performed meta-analysis. It is thought that this meta-analysis was the first to evaluate clinical efficacy of LPD for the treatment of pancreatic cancer with long-term survival outcomes. It was found that there was no significant difference in the 5-year overall survival (HR: 0.97, 95% CI: 0.82–1.15, *p* = 0.76). In addition, LPD resulted in a higher rate of R0 resection, more harvested lymph nodes, shorter hospital stays, and less estimated blood loss, concluding that LPD is not inferior to OPD with respect to long-term oncologic outcomes, as well as better short-term surgical outcomes in patients with pancreatic cancer.

Chen et al. [82] also evaluated 1507 patients from 6 comparative cohort studies, comparing LPD with OPD for pancreatic cancer. Similar short-term oncologic outcomes were identified, such as lymph nodes harvested, R0 rate, number of positive lymph nodes, adjuvant treatment, and time to adjuvant treatment. Interestingly, in spite of comparable 1-year and 2-year survival, the following 3-year (OR 1.50, *p* = 0.007), 4-year (OR 1.73, *p* = 0.04), and 5-year survivals (OR 2.11, *p* = 0.001) were significantly longer in LPD group.

Yin et al. [83] identified 6 studies including 9144 pancreatic cancer patients and evaluated short-term and long-term oncological outcomes. They noted that fewer postoperative complications (*p* = 0.005), better trend of performance in R0 resection (*p* = 0.07), retrieved number of lymph nodes (*p* = 0.07), and comparable long-term survival (*p* = 0.49) were associated with LPD, concluding that LPD can be a suitable alternative to OPD in selected PDAC patients with respect to both surgical and oncological outcomes.

In addition, there are several articles analyzing the national cancer data base (NCDB) to overview the safety and effectiveness of LPD in treating pancreatic cancer. Sharpe et al. (study period: 2010–2011) [84] showed that, among the 4421 patients, a very limited number of patients (384 patients, 9%) underwent LPD for pancreatic cancer, and about one third of cases (118 patients, 30%) were

performed in high volume centers (5 institutions, 3.8%). They demonstrated that LPD was equivalent to OPD in length of stay, R0-resection, lymph node count, and readmission rate, however a higher 30-day mortality rate in LPD group was noted in lower volume centers (7.5% vs. 3.4%, *p* = 0.003), raising concerns about the safety of LPD in treating pancreatic cancer due to a surmountable learning curve for the procedure. Kantor et al. (study period: 2010–2013) [85] analyzed 8213 patients with pancreatic cancer to show short-term and long-term oncologic outcomes of LPD and OPD. They found that LPD (828 patients, 10%) provides comparable short-term oncologic and long-term overall survival outcomes with OPD (20.7 months vs. 20.9 months, *p* = 0.68). In addition, decreased incidence of prolonged length of hospital stay (OR = 0.79), decreased rate of readmission and decrease in incidence of delay to adjuvant chemotherapy for LPD (OR 0.71, *p* = 0.11) were noted in LPD, suggesting a trend towards accelerated recovery. Higher 30-day mortality was still noted in LPD performed by lower volume centers (<20 LPDs, 5.6% vs. 0%, *p* < 0.01), however no significant differences of 30-day and 90-day mortality in high volume centers (≥20 LPDs) between LPD and OPD (0% vs. 1.3%, *p* = 0.08, and 0.6% vs. 3.4% *p* = 0.08, respectively). Chapman et al. (study period: 2010–2013) [86] evaluated the safety of LPD in elderly patients (≥75 years old, 1768 patients). Although it was found that more than 60% of the patients with LPD was performed in low volume centers (<5 LPDs), no significant differences were found in R0-status, number of lymph nodes examined, lymph node, status, receipt of adjuvant chemotherapy, days to initiation of adjuvant chemotherapy, readmission rates, or 30-day mortality between two groups (*p* > 0.05). Ninety-day mortality was significantly lower in LP (7.2 vs. 12.2%, *p* = 0.049). It was observed that there was a trend towards improved OS (HR = 0.85, *p* > 0.05) in the LPD group compared to the OPD group after adjusting for patient and tumor-related characteristics. Lastly, Torphy et al. (study period: 2010–2015) [87] also compared short-term and oncologic outcome of minimally invasive PD (MIPD) with those of OPD across low and high-volume centers. Among the patients with pancreatic cancer who underwent PD, 3754 patients (17.1%) were found to be performed minimally invasively. It was found that patients with MIPD for pancreatic cancer were less likely to stay in the hospital (OR, 0.75; 95% CI, 0.68–0.82). Thirty-day mortality, 90-day mortality, unplanned readmissions, margins, lymph nodes harvested, and receipt of adjuvant chemotherapy were equivalent between two groups.

Table 6 shows the recent significant articles comparing LPD and OPD in treating pancreatic cancer. When looking at the individual data, they are all pointing out that LPD can provide equivalent or superior short-term oncologic outcomes with comparable long-term survival outcome, suggesting that LPD is technically feasible and even oncologically safe in treating pancreatic cancer. It is interesting to note that every reported article demonstrates a smaller amount of intraoperative blood loss comparing with OPD. This observation is very important to improve the long-term oncologic outcomes of resected pancreatic cancer, because a lower volume of intraoperative estimated blood loss can lead to a lower chance of unnecessary intraoperative blood transfusion. Potential deleterious effects of blood transfusion on oncologic outcomes have been explained by several hypothesized mechanisms, primarily via the induction of immunosuppression [88,89]. In several studies, intraoperative transfusion was found to be one of the independent prognostic factors in resected pancreatic cancer [90–92]. In addition, a recent meta-analysis [35] to evaluate potential relationship between perioperative blood transfusion and prognosis of pancreatic cancer surgery also demonstrated detrimental effect of blood transfusion on survival in univariate (68.4%, 13 out of 19 studies) and multivariate analysis (47.4%, 9 out of 19 studies) respectively, showing overall blood transfusion associated with pancreatic cancer surgery can be related to shorter overall survival (pooled odds ratio 2.43, 95% confidence interval 1.90–3.10). Considering poor long-term oncologic outcomes of pancreatic cancer, intraoperative transfusion must be an attractive issue for pancreatic surgeons, because it is thought to be a surgeon-controllable factor by reducing intraoperative blood loss. In the near future, appropriate transfusion guidelines including strict transfusion threshold and active investigation on alternative to allogenic blood transfusion are necessary in managing pancreatic cancer patients.


*Cancers* **2020**, *12*, 3430

However, all these articles showing comparable oncologic outcomes between LPD and OPD are based on retrospective studies, where unavoidable selection bias should be always concerned. Therefore, the advantages of laparoscopic PD on short-term surgical outcomes and comparable long-term survival rates should be regarded with careful interpretation.

However, unfortunately there is no randomized control trial (RCT) comparing LPD and OPD for pancreatic cancer. All conclusions were based on retrospective observational studies. Therefore, the interpretation of above these articles should be careful because there are unavoidable limitations, such as small sample size, selection bias, learning-curve issue, and heterogeneity in reported data.

Which approach, laparoscopic vs. open, is better regarding surgical approach to pancreatic cancer? An RCT is still needed to actually elucidate the true oncologic value of LPD, and to suggest standard care in treating pancreatic cancer. However, an RCT regarding this issue is not that easy to conduct because most pancreatic cancers are found to be at an advanced stage at diagnosis, and the surgical technique of LPD is not fully matured in the present surgical society. In fact, three RCTs comparing LPD and OPD were reported [13–15], but their study population included not only pancreatic cancer but also other periampullary cancer, benign and low-grade malignant tumors of the pancreas. Overall, they demonstrated no statistical differences, except less blood loss in LPD, and shorter operative time in OPD.

However, when looking at the last study recently performed by Hilst et al. [15], nine surgeons who had experienced more than 20 cases of LPD or OPD were involved. This clinical trial was prematurely terminated due to safety concerns of LPD. This phenomenon is thought to be a part of reflection to previous NCDB-based studies showing LPDs done in low volume centers were associated with higher mortality rate [84,85]. Taking the tumor biology of pancreatic cancer and its clinical presentations into consideration, associated pancreatitis, cholangitis, and potential risk of vascular involvements usually make laparoscopic oncologic dissection much more difficult and even dangerous. Therefore, LPD for treating pancreatic cancer should be a more challenging issue. All recent studies agree with the fact that many cases are required to overcome the learning curve for safe LPDs [16,17]. Therefore, under reasonable inclusion criteria, it is highly demanded that expert surgeons with good experience in both LPD and OPD should collaborate for a well-designed RCT to answer this question (ClinicalTrials.gov Identifier: NCT03870698). For a while, the currently accumulated expert surgeons' experiences might be regarded as the highest level of evidence that we can take in current clinical situation of pancreatic cancer.

#### **4. Proposal Potential Indications**

Previously, we suggested a model for determining the indications of minimally invasive radical distal pancreatectomy for left-sided pancreatic cancer [7]. A similar approach will be possible in LPD for pancreatic cancer. Technical feasibility, procedural safety (or surgical risk), and surgical extent for margin-negative resection (oncologic clearance) should be considered in defining potential indication of LPD. A surgeon's capacity of technical feasibility should cover the surgical extent requiring for curative resection. Figure 4 shows the dynamic relationships between these factors.

As the surgical extent for curative resection increases, the following will happen: First, the potential surgical risk may increase. Second, advanced surgical technique is highly required to ensure surgical extent for obtaining margin-negative resection. However, at some point, an appropriate surgical procedure for curative resection cannot be maintained by the laparoscopic approach due to surgeons' own technical issues and patients' co-morbidities. Therefore, there must be an optimum surgical extent that can be obtained by individual surgeons' own surgical techniques for margin-negative resection.

From that point of view, three types of surgical extent of LPD can be available. *Type 0* LPD is PD with standard dissection. It does not require any type of combined vascular resection. *Type I* LPD requires combined venous vascular resection. In this surgical extent, most cases need tangential (wedge) resection of PV, or SMV with primary repair (*Type Ia* LPD). *Type Ib* LPD requires more complicated procedures for margin-negative resection, such as tangential resection with patch repair, segmental

resection with end-to-end anastomosis or reconstruction using artificial graft (venous resection 2). In selected cases, combined arterial resection (*Type II* LPD) will be required, but quite limited [93].

For example, (A) surgeons' technique for LPD is not enough to achieve curative resection in resectable pancreatic cancer. This group of surgeons need to do open PD for curative resection of the pancreatic cancer. However, (B) surgeons' technique for LPD is good enough for obtaining margin-negative radical resection. They may perform LPD in well-selected cases expecting no combined vascular resection (Type 0 LPD). Only a few surgeons can perform LPD even with combined venous vascular resection (tangential or segmental) or even combined arterial resection in well-selected patients (Type I and II LPD). Therefore, considering the present technical feasibility to maintain patient's safety, and margin-negative curative pancreatectomy, tumor conditions that could be removed by standard PD without combined vascular resection (Type 0 LPD) will be the primary indication for LPD. Others (Type I and II) can be performed in selected cases by only expert surgeons. *Cancers* **2020**, *12*, x 12 of 20 pancreatic cancer. However, (B) surgeons' technique for LPD is good enough for obtaining marginnegative radical resection. They may perform LPD in well-selected cases expecting no combined vascular resection (Type 0 LPD). Only a few surgeons can perform LPD even with combined venous vascular resection (tangential or segmental) or even combined arterial resection in well-selected patients (Type I and II LPD). Therefore, considering the present technical feasibility to maintain patient's safety, and margin-negative curative pancreatectomy, tumor conditions that could be removed by standard PD without combined vascular resection (Type 0 LPD) will be the primary

When looking at the anatomic relationship between pancreatic cancer and major vascular structures at diagnostic stage, anatomically "*resectable*" pancreatic cancer includes potential candidates who can undergo LPD for curative intent in clinical practice [94]. However, unexpected involvement of adjacent venous systems, such as superior mesenteric vein (SMV), or portal vein (PV) because of severe pancreatitis or tumor invasion, can be encountered during LPD for resectable pancreatic cancers, and this situation might be necessary for combined vascular resection. Only expert surgeons are responsible for this advanced stage of the pancreatic cancer. These cases mostly will result in elective conversion to open for procedural safety and curative resection [95]. indication for LPD. Others (Type I and II) can be performed in selected cases by only expert surgeons. When looking at the anatomic relationship between pancreatic cancer and major vascular structures at diagnostic stage, anatomically "*resectable*" pancreatic cancer includes potential candidates who can undergo LPD for curative intent in clinical practice [94]. However, unexpected involvement of adjacent venous systems, such as superior mesenteric vein (SMV), or portal vein (PV) because of severe pancreatitis or tumor invasion, can be encountered during LPD for resectable pancreatic cancers, and this situation might be necessary for combined vascular resection. Only expert surgeons are responsible for this advanced stage of the pancreatic cancer. These cases mostly will result in elective conversion to open for procedural safety and curative resection [95].

**Figure 4.** Model for determining surgical indication of LPD for pancreatic cancer. Regardless of laparoscopic or open PD for pancreatic cancer, surgical approach should be allowed only when surgeons' technical feasibility can obtain the appropriate surgical extent for margin-negative resection. Therefore, indication of LPD can vary according to surgeons' techniques and disease extent. OPD will be recommended in surgeon (**A**). Type 0 LPD can be done in surgeon (**B**). Type Ia LPD and Type Ib LPD can be allowed for surgeon (**C**) and surgeon (**D**), respectively. Anatomically *resectable* pancreatic cancer with intact fat plane between pancreas and major vascular structures is thought to be the ideal tumor conditions for LPD (Tumor conditions controlled by Type 0 LPD). Note: This author follows Fortner's initial classification of regional pancreatectomy [96]. **Figure 4.** Model for determining surgical indication of LPD for pancreatic cancer. Regardless of laparoscopic or open PD for pancreatic cancer, surgical approach should be allowed only when surgeons' technical feasibility can obtain the appropriate surgical extent for margin-negative resection. Therefore, indication of LPD can vary according to surgeons' techniques and disease extent. OPD will be recommended in surgeon (**A**). Type 0 LPD can be done in surgeon (**B**). Type Ia LPD and Type Ib LPD can be allowed for surgeon (**C**) and surgeon (**D**), respectively. Anatomically *resectable* pancreatic cancer with intact fat plane between pancreas and major vascular structures is thought to be the ideal tumor conditions for LPD (Tumor conditions controlled by Type 0 LPD). Note: This author follows Fortner's initial classification of regional pancreatectomy [96].

candidates for LPD are certain to increase

Therefore, in the beginning stage, it is thought that only a small proportion of resectable

Therefore, in the beginning stage, it is thought that only a small proportion of resectable pancreatic cancer without contact with the SMV-PV complex could be a potential indication for LPD (Type 0 LPD) for generalizing concept of LPD for pancreatic cancer (Figure 5). However, in the near future, with the advance of a diagnostic strategy for early detection and surgical techniques, potential candidates for LPD are certain to increase *Cancers* **2020**, *12*, x 13 of 20

**Figure 5.** A case with extremely appropriate pancreatic cancer for LPD. A case of early pancreatic cancer in a 71-year-old, male patient, incidental finding of pancreatic duct dilatation during routine medical check-up. Preoperative image studies showed less than 1 cm sized pancreatic mass with secondary pancreatic duct dilatation (**a**,**b**). Type 0 LPD was performed on 24 October 2019, and discharged 9 days postoperatively. Pathological examination reported 3.0 mm-sized pancreatic ductal adenocarcinoma with good differentiation (×100, **c**). resected margins were free from malignant cells (safety margin ≥ 1cm). No lymph node metastasis was noted among 23 retrieved lymph nodes. **Figure 5.** A case with extremely appropriate pancreatic cancer for LPD. A case of early pancreatic cancer in a 71-year-old, male patient, incidental finding of pancreatic duct dilatation during routine medical check-up. Preoperative image studies showed less than 1 cm sized pancreatic mass with secondary pancreatic duct dilatation (**a**,**b**). Type 0 LPD was performed on 24 October 2019, and discharged 9 days postoperatively. Pathological examination reported 3.0 mm-sized pancreatic ductal adenocarcinoma with good differentiation (×100, **c**). resected margins were free from malignant cells (safety margin ≥ 1 cm). No lymph node metastasis was noted among 23 retrieved lymph nodes.

#### **5. Conclusions 5. Conclusions**

fulfil these two goals.

2015-0053.

It is worthy to emphasize that margin-negative radical pancreatectomy is the essential condition to obtain long-term survival of patients with pancreatic cancer. With the investigation for early diagnosis, introduction of potent chemotherapeutic agents, application of neoadjuvant chemotherapy, advancement of open and laparoscopic surgical techniques, mature perioperative management, and improved patients' general conditions, survival of resected pancreatic cancer is expected to be further improved. OPD will still be mainstream in treating resectable pancreatic cancer. Particularly, OPD with anatomy-driven extended dissection will play significant role in treating advanced pancreatic cancer following neoadjuvant chemotherapy, which harbors the potential risk of combined major venous or arterial resection. On the other hand, the technical and oncological feasibility of LPD for pancreatic cancer is still It is worthy to emphasize that margin-negative radical pancreatectomy is the essential condition to obtain long-term survival of patients with pancreatic cancer. With the investigation for early diagnosis, introduction of potent chemotherapeutic agents, application of neoadjuvant chemotherapy, advancement of open and laparoscopic surgical techniques, mature perioperative management, and improved patients' general conditions, survival of resected pancreatic cancer is expected to be further improved. OPD will still be mainstream in treating resectable pancreatic cancer. Particularly, OPD with anatomy-driven extended dissection will play significant role in treating advanced pancreatic cancer following neoadjuvant chemotherapy, which harbors the potential risk of combined major venous or arterial resection.

controversial. Considering difficult clinical circumstances to perform RCT, the currently reported experiences are thought to be the highest level of evidence that we can consider in managing pancreatic cancer. According to the literature, it can be carefully concluded that LPD is thought to be a good alternative strategy in managing well-selected resectable pancreatic cancer. LPD with combined vascular resection is also feasible, but only expert surgeons should handle these challenging cases. On the other hand, the technical and oncological feasibility of LPD for pancreatic cancer is still controversial. Considering difficult clinical circumstances to perform RCT, the currently reported experiences are thought to be the highest level of evidence that we can consider in managing pancreatic cancer. According to the literature, it can be carefully concluded that LPD is thought to be a good alternative strategy in managing well-selected resectable pancreatic cancer. LPD with combined vascular resection is also feasible, but only expert surgeons should handle these challenging cases.

In spite of potential advantages of the laparoscopic approach, it provides surgeons with fundamental limitations during surgical procedures, such as 2-D operative view, fulcrum-effect, limited motion of the instruments, attenuated touch sensation, and enhancing tremor. Theoretically, the robotic surgical system was introduced to overcome these hurdles. However, it is also true that many laparoscopic surgeons had already overcome these limitations. Therefore, surgical approach should be determined according to the surgeon's expertise, patients' general condition, and tumor biology. As long as surgical principles for pancreatic cancer are kept in mind, open, laparoscopic or robotic approach to PD for pancreatic cancer are expected to provide meaningful short- and longterm oncologic outcomes of resected pancreatic cancer. Finally, in application of minimally invasive pancreatectomy for pancreatic cancer, patients' In spite of potential advantages of the laparoscopic approach, it provides surgeons with fundamental limitations during surgical procedures, such as 2-D operative view, fulcrum-effect, limited motion of the instruments, attenuated touch sensation, and enhancing tremor. Theoretically, the robotic surgical system was introduced to overcome these hurdles. However, it is also true that many laparoscopic surgeons had already overcome these limitations. Therefore, surgical approach should be determined according to the surgeon's expertise, patients' general condition, and tumor biology. As long as surgical principles for pancreatic cancer are kept in mind, open, laparoscopic or robotic approach to PD for pancreatic cancer are expected to provide meaningful short- and long-term oncologic outcomes of resected pancreatic cancer.

safety and principles of surgical oncology should be kept in mind [97]. In fact, these two principles should be considered not only in LPD but also in OPD. Whether PD is performed by laparoscopic or Finally, in application of minimally invasive pancreatectomy for pancreatic cancer, patients' safety and principles of surgical oncology should be kept in mind [97]. In fact, these two principles should

be considered not only in LPD but also in OPD. Whether PD is performed by laparoscopic or open approach, surgical approach should be determined by a surgeon's own technical expertise to fulfil these two goals.

**Funding:** This research was funded by a Faculty Research Grant of Yonsei University College of Medicine for 6-2015-0053.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Achieving 'Marginal Gains' to Optimise Outcomes in Resectable Pancreatic Cancer**

**Sarah Powell-Brett 1,\*, Rupaly Pande 1,† and Keith J. Roberts 1,2**

	- **\*** Correspondence: sarah.powell-brett1@nhs.net; Tel.: +44-(0)788-775-4856
	- † Joint 1st authors.

**Simple Summary:** Improving outcomes in pancreatic cancer is achievable through the accumulation of marginal gains. There exists evidence of variation and undertreatment in many areas of the care pathway. By fully realising the existing opportunities, there is the potential for immediate improvements in outcomes and quality of life.

**Abstract:** Improving outcomes among patients with resectable pancreatic cancer is one of the greatest challenges of modern medicine. Major improvements in survival will result from the development of novel therapies. However, optimising existing pathways, so that patients realise benefits of already proven treatments, presents a clear opportunity to improve outcomes in the short term. This narrative review will focus on treatments and interventions where there is a clear evidence base to improve outcomes in pancreatic cancer, and where there is also evidence of variation and under-treatment. Avoidance of preoperative biliary drainage, treatment of pancreatic exocrine insufficiency, prehabiliation and enhanced recovery after surgery, reducing perioperative complications, optimising opportunities for elderly patients to receive therapy, optimising adjuvant chemotherapy and regular surveillance after surgery are some of the strategies discussed. Each treatment or pathway change represents an opportunity for marginal gain. Accumulation of marginal gains can result in considerable benefit to patients. Given that these interventions already have evidence base, they can be realised quickly and economically.

**Keywords:** pancreatic cancer; pancreatic exocrine insufficiency; adjuvant chemotherapy; biliary drainage; prehabilitation; ERAS

#### **1. Introduction**

Pancreatic cancer is projected to become the second leading cause of all cancer-related deaths by 2030 [1,2]. Although surgical resection is the foundation of 'resectable' pancreatic cancer management, alone it is associated with a less than 10% chance of cure [3]. High rates of perioperative morbidity and mortality, slow or poorly organised pathways to surgery, suboptimal preoperative management of jaundice, suboptimal use of (neo)adjuvant therapy and failure to address malnutrition all contribute to poor outcomes. It is therefore unsurprising that there can be a negative attitude and feelings of nihilism when considering the outlook for these patients.

Wide variations in care and a lack of standardised practice, however, offer an easy way to improve outcomes in the near future. Pathways to expedite surgery, prescribing of enzyme therapy, access to adjuvant chemotherapy and tackling frailty are just a few examples of how care has the potential to be optimised. Optimising an individual aspect of care represents an opportunity for a marginal gain. Individually, each gain realised may be relatively minor when observed across an entire patient cohort, but when multiple, the effect of aggregated marginal gains can be considerable. The aggregation of marginal

**Citation:** Powell-Brett, S.; Pande, R.; Roberts, K.J. Achieving 'Marginal Gains' to Optimise Outcomes in Resectable Pancreatic Cancer. *Cancers* **2021**, *13*, 1669. https://doi.org/ 10.3390/cancers13071669

Academic Editor: Tsutomu Fujii

Received: 4 February 2021 Accepted: 24 March 2021 Published: 1 April 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

gains was popularised in elite level cycling where its success was clear to see at Olympic level and major events such as the Tour de France. The principals have been adopted in healthcare among various patient populations from cancer surgery, stroke recovery, prehabilitation, cardiac surgery and anaesthesia [4–6]. This narrative review highlights common failings in the care of pancreatic cancer patients and describes where gains can be made. The aggregation of these marginal gains could improve outcomes and experience for a great many patients with resectable pancreatic cancer.

#### **2. Pre-Operative Pathways**

At presentation, resectable pancreatic cancer is in an exponential phase of growth and most primary tumours harbour cells that can metastasise [7]. Enhancing pre-operative pathways to ensure that patients are treated as quickly as possible and their functional status is optimised is essential to enabling the delivery of the highest standard of care. This section will focus on avoidance of pre-operative biliary drainage, prehabilitation, and optimisation of nutritional status.

#### *2.1. Preoperative Biliary Drainage in Resectable Pancreatic Cancer*

The majority of patients with resectable pancreatic cancer present with jaundice. Historically reluctance to operate in the presence of jaundice was related to concerns over renal, cardiac and liver dysfunction, coagulopathy and, at the time, high rates of perioperative morbidity and mortality associated with pancreatic surgery in the absence of jaundice [8].

Hence, in theory, relief of jaundice by preoperative biliary drainage (PBD) is perceived to improve these disturbances and prevent postoperative complications, though in practice, the role and propriety of PBD must be challenged.

The concept of correcting jaundice prior to resection was introduced by A.O. Whipple through staged PD, initially a cholecystogastrostomy to relieve jaundice followed by resection once the jaundice was within 'safe' limits. This further developed into nonoperative approaches through percutaneous transhepatic cholangiopathy (PTC) and biliary drainage in the 1960s and later followed by endoscopic retrograde cholangiopancreatography (ERCP) in the 1970s [9,10].

For many decades ERCP served both diagnostic and therapeutic purposes. However, the sensitivity and specificity of CT to diagnose pancreatic cancer has made ERCP, as a diagnostic tool, obsolete.

The main drawback of PBD is the associated rate of complications. These complications impair quality of life, can delay or prevent future surgery and are occasionally fatal. PBD itself requires resource and invariably delays treatment pathways [11,12].

The DROP trial randomised patients to upfront early surgery within 7 days or PBDand plastic stent and delayed surgery between 4–6 weeks [13]. This landmark RCT showed the rate of serious complications at 120 days postoperatively to be far higher in the PBD group (74 vs. 39%, *p* < 0.001). This was largely due to drainage-related complications, cholangitis (26 vs. 2%), pancreatitis (7 vs. 0%) along with the need for change of stent in 30% of patients. There are also higher rates of perioperative infections, associated with changes in the biliary microbiome related to PBD [14,15] It is therefore advised that intraoperative bile cultures are taken and appropriate antibiotic administered to address this higher rate of complications [16].

Self-expanding metal stents (SEMS) have gained in popularity over plastic stents based on significantly greater patency and reduction in infectious complications [13,17,18]. The experience of patients undergoing PBD with fully covered SEMS, with a prospective study which mirrored the protocol of the DROP trial, superimposed the complication rates of PBD-SEMS over the cohorts within the DROP trial [19]. Though SEMS were associated with lower rates of complications than plastic stents, the rate remained significantly greater than upfront surgery (51 vs. 39%). Subsequent meta-analysis confirms advantages of metal over plastic stents with reduced rates of need for endoscopic reintervention (OR0.3),

preoperative complications (OR 0.42) and cholangitis (OR 0.09) [20]. However, pancreatitis is more common (OR 3.6) and, though not reported, post hoc analysis of resection rates demonstrates a lower resection rate among patients with metal stents (plastic, 147/190 vs. metal 93/139; Chi square *p* = 0.04) [19,20]. A more recent network meta-analysis demonstrated a remarkably high rate of complications with plastic stents (38–93%), a much lower rate with SEMS (0–15%) whilst data regarding outcomes from percutaneous drainage is not at extensively reported (31%) [21]. This study also concluded that avoidance of PBD was associated with the best outcome.

The association between PBD and reduced resection rates was not observed in individual randomised trials. However, it is logical to associate complications such as pancreatitis or pathway delays with a reduction in resection rate. Many high-volume centres have reported that increasing time to surgery reduces resection rates, data confirmed within a recent systematic review and meta-analysis [22–25]. It may be that this higher resection rate and avoidance of complications translates into increased survival benefit, when analysed on an intention to treat basis [26]. This is controversial as some authors have associated hyperbilirbuinaemia at the time of resection with worse cancer outcomes [27].

Given the many benefits of avoiding PBD, the National Institute for Health and Care Excellence (NICE) in the United Kingdom, now recommend upfront surgery without PBD where possible [28].

What remains to be defined is a safe upper limit of bilirubin in terms of safety and oncologic outcomes and why PBD remains so widely used despite the evidence of harm. Associated venous resection appears safe in the presence of jaundice, even when levels of bilirubin exceed 300 µmol/L (>17.5 mg/dL) [29]. It may be assumed that elderly patients are safer to undergo PBD and surgery rather than upfront surgery but elderly patients are less likely to tolerate complications of PBD and remain on a surgical pathway than younger patients.

Numerous studies have established seemingly arbitrary upper limits of bilirubin at which PBD is indicated [27,30–35]. A common threshold of >250 umol/L is used and cited by The Guidelines for Perioperative Care for Pancreaticoduodenectomy: Enhanced Recovery After Surgery (ERAS®) Society Recommendations [36]; Both Sauvanet and Li et al. reported higher complication rates when surgery was performed with a bilirubin of >300 µmol/L [27,37]. However, contrary to this, van der Gaag, quoted a bilirubin level of ≤300 µmol/L at surgery and demonstrated significantly lower infective complications, whilst Pamecha et al. showed that a bilirubin level of ≥15 mg/dL (≥265 µmol/L) was not an independent risk factor for complications [13,38,39].

As stated above, it is the authors experience that surgery, even with venous resection, can be performed with no difference in complications with bilirubin in excess of 300 µmol/L. To summarise, there is no clear data that defines an upper limit of bilirubin at which PBD is indicated. The authors recommend an approach with prospective evaluation of outcomes and a step wise increase in threshold of bilirubin at which to undertake PBD. Variation in the rate of increase of bilirubin, the difficulty/time needed to complete staging tests and other factors such as renal function or evidence of biliary sepsis mean that there is unlikely to be a single value which can be applied to all patients.

It is useful to consider when PBD is indicated. Jaundiced patients undergoing neoadjuvant therapy clearly require PBD; if there are major diagnostic delays, though an assessment of risk of cancer progression must be weighed up against potential risks of surgery in such cases or if the patient is too frail to undergo surgery. Such patients, however, in our experience rarely improve significantly after PBD. Cholangitis has been considered a contraindication for PBD, though in our practice we frequently employ external biliary drainage, antibiotics, fluid replacement therapy and surgery within the same week if markers of infection are improving. This has controlled sepsis, avoided complications of PBD and kept patients within an early surgery pathway. See Table 1. For a summary of potential indications for PBD and Table 2 for optimizing care should PBD be undertaken.


**Table 1.** Absolute and relative indications for Percutaneous biliary drainage (PBD).

**Table 2.** Tips for optimising care in the event of PBD requirement.


Malnutrition associated with obstructive jaundice and its effects on outcomes following surgery is a further area of controversy. In studies by Padillo et al., malnutrition was more commonly associated with patients older than 68 years and those with high levels of bilirubin with the suggestion that PBD should be considered to allow alleviation of these modifiable factors preoperatively [40,41]. However, evidence for the optimal duration of PBD to improve nutritional status is experimental and its role in malnutrition must be considered in the broader context of pancreatic exocrine insufficiency [42–45]. Exocrine insufficiency is prevalent at diagnosis, and although not the sole cause of malnutrition, it is a major driver of malnutrition and is poorly treated; thus, delays to surgery can exacerbate malnutrition if PEI is not addressed. A balance must be struck between early surgery and better treatment of PEI versus correcting jaundice and improving nutrition. A proposed optimal pathway would be to provide PERT for all patients and provide early surgery where

possible; where a patient has a poor performance status with malnutrition, a pathway of PBD, PERT, and dietician input would be advocated.

#### *2.2. Pancreatic Exocrine Insufficiency and Overcoming Malnutrition to Improve Outcomes*

Pancreatic exocrine insufficiency (PEI) and pancreatic enzyme replacement therapy (PERT) should be considered at all stages of management of pancreatic cancer, it is included in this section on pre-operative pathways as correct treatment is essential as early in the care pathway as possible. PEI is far from trivial and should be considered as organ failure. Untreated failure of other organ systems can be rapidly fatal (for example, renal, cardiac or respiratory failure or even diabetes) and withholding treatment in these settings is reserved only for those patients who are on end-of-life pathways. Consequently, it is remarkable that PEI is underdiagnosed and undertreated. *In the authors view, this single issue represents the simplest pathway improvement in this review and also the intervention with the chance to achieve most gain.*

Addressing pancreatic exocrine insufficiency (PEI) can have an effect upon survival as strong as that as surgery or chemotherapy and yet many patients remains untreated [46]. A 2016 systematic review of PEI demonstrated a pre-operative prevalence of 44% and a post-operative prevalence of 74% (for those undergoing pancreatico-duodenectomy for malignancy). This is likely a conservative estimate owing to the frequent use of faecal elastase (FE-1) to diagnose PEI (FE-1 has been shown to underestimate PEI following resection) [47,48]. Furthermore, there is a gradual reduction of exocrine function at a median value of 10% per month [49]. With the gold standard of testing and longer term follow up, Lemaire et al. found the post-operative incidence of PEI to be 94% [50]. In resectable disease the mechanisms underlying PEI are complex and multifactorial. Almost all physiologic control is lost after pancreatoduodenectomy, and together with other factors result in insufficient enzymes arriving at the wrong time, to the wrong place and at the wrong pH for effective function [51,52] (see Figure 1).

**Figure 1.** Factors contributing to PEI following pancreatico-duodenectomy. \*CCK = cholecystokinin.

Symptoms of PEI include pain, bloating, frequency, urgency, diarrhoea, fatty stool, flatulence, loss of appetite, nausea and vomiting and are physically and mentally distressing. The 'classical' sign of steathorrhoea is often absent, either because PEI is not severe enough or because the patient may have unconsciously adopted behaviours to avoid fat intake [53–55]. PEI is frequently untreated or undertreated worldwide. Studies from the UK, Europe and Australia demonstrate that only a minority of patients receive PERT [56–59]. Reasons for under prescribing are multifactorial and may relate to the lack of an acceptable or accurate routine diagnostic test that can yield results in near real time [47,60–62]. Confusing symptoms of cancer, weight loss and abdominal discomfort, further complicate diagnosis.

The consequences of PEI must not be underestimated, these include: weight loss, malnutrition, micronutrient deficiency, cardiovascular events, osteoporosis, fractures and sarcopenia [63]. The effect of PEI on operative outcomes is considerable, being associated with higher rates of post-operative complications, longer hospital stays and increased costs [64–67].

Pancreatic enzyme replacement therapy (PERT) is cheap, mitigates against weight loss and improves quality of life. The most compelling argument for PERT is emerging evidence that it improves survival [46,58,68–71]. Given the failings of diagnostic tests for PEI and the evident benefits of PERT, it is recommended for all patients with pancreatic cancer by NICE in the United Kingdom [28].

A healthy pancreas is estimated to produce 900,000 United States Pharmacopeia (USP) of lipase in response to a meal. Sufficient absorption of fat can be maintained at around 10% of normal capacity, there is thus a need for around 90,000 USP per meal. Given that in the majority of pancreatic disease some function remains an appropriate starting dose in resected pancreatic cancer is 75,000 USP with a main meal and 25,000 with a snack [69], most effective when given across the course of, or just after, a meal rather than before [72]. Co-prescription of a proton pump inhibitor is often required after pancreatoduodenectomy as a failure to neutralise gastric acid leads to enzymes remaining inactive [73].

Pre-operative malnutrition is associated with significantly poorer post-operative outcomes for pancreatic resection, addressing this extends beyond just the prescribing of PERT [74,75]. The international Study Group on Pancreatic Surgery (ISGPS) released a position paper on nutritional support and therapy in pancreatic surgery, this emphasizes the importance of pre-operative assessment of nutritional status and recommends nutritional supplements in those who have, or are at risk of developing, moderate malnutrition [76]. Those with, or at risk of severe malnutrition may benefit from formal nutritional support with enteral or parenteral feeding [76]. Regular dietician input to asses response, compliance, diet, diabetic optimization, and the potential need for nutritional supplements is required [77].

Although outside the scope of this article to describe in detail, immunonutrition is worthy of note. Immunonutrition aims to influence the systemic immune system using nutritional supplements with immune modulating contents such as arginine, omega-3 fatty acid and RNA. This is not yet part of routine clinical practice, however, there is evidence that initiation in the pre-operative period could improve post-operative outcomes [78–81]. The most recent systematic review and meta-analysis of immunonutrition in pancreatic resection, concluded that immunonutrition reduces infectious complications (especially wound infection) and length of stay [82].

#### *2.3. Benefits of and Access to Surgical Resection of Pancreatic Cancer in the Elderly*

Benefits of resection do not diminish with increasing age [83,84]. Yet, many elderly patients are considered too frail for surgery and there is significant age-related disparity in access to surgery. Ageism is a major problem faced by elderly cancer patients; elderly patients without comorbidity are less likely to receive cancer therapy than younger patients with comorbidity. Advanced age, in the absence of comorbidity, is mistakenly considered a more significant barrier to surgery than comorbidity [85]. Elderly patients are less likely to undergo standard resection, less likely to undergo resection with concomitant venous resection and less likely to achieve negative margins [86–88]. An American population based study of over 45,000 patients with pancreatic adenocarcinoma observed that the rate of surgery decreased with increasing age: 21% of those under 50, 19% between the age of 50 and 70 and only 13% of those over 70 received surgery [87].

Concerns over risk and perceived lack of benefit of surgery are prohibitive and account for the discrepancies in care. However, a systematic review of surgery among elderly patients with pancreatic cancer demonstrated that over time, perioperative mortality has improved for elderly patients following pancreatic resection when compared to non-elderly patients, with mortality preceding the year 2000 being significantly higher in elderly patients, but similar from 2000 onwards [86]. Major surgical complications (post-operative pancreatic fistula, delayed gastric emptying, post pancreatectomy haemorrhage and surgical site infections) were similar between the elderly and the non-elderly; however, respiratory complications did occur more frequently in the elderly population. Prehabilitation can improve physical functioning and prevent deterioration among patients whilst waiting for cancer surgery [89–94]. The benefits of surgery are not diminished by age and therefore older patients with appropriate performance status should not be denied access based on chronological age alone. Treatment decisions for the elderly should be made in a multidisciplinary setting, should ideally include the use of a tool such as the comprehensive geriatric assessment and the input of a geriatrician to avoid discrepancies in treatment based on chronological age [95,96].

#### *2.4. Prehabilitation*

Resection is the only curative option for pancreatic cancer, however, the majority of patients with a new diagnosis are not suitable for operative management [97]. Baseline functional status influences receipt of curative resection, receipt of adjuvant chemotherapy, the rate and severity of post-operative complications and long term quality of life [89,98–100]. Functional status is often poor in patients with pancreatic cancer, this is contributed to by older age at diagnosis, pre-operative sarcopenia, malnutrition and obstructive jaundice [101]. Prehabilitation refers to any interventions, prior to definitive treatment that are aimed at improving patient health and lifestyle [102]. The concept centres around pre-operative conditioning to improve nutritional status and aerobic capacity. Studies of prehabilitation in other types of major surgery have suggested that these programmes can improve both access to definitive treatment, and post-operative outcomes [103] (See Table 3). A recent trial randomised patients to standard of care versus standard of care plus prehabilitation in patients having major abdominal surgery found that the prehabilitation cohort had improved aerobic capacity and a reduction in 30-day readmissions [104]. Unfortunately, systematic reviews and meta-analysis have determined that although prehabilitation programmes can potentially improve surgical outcomes, the evidence is weak, this is most likely do to the variation in prehabilitation regimens and study heterogeneity [102,103,105–108].

The majority of evidence for prehabilitation is limited to those undergoing colorectal resection, hepatic resection or major cardio-thoracic surgery, few studies look specifically at the impact of prehabilitation in pancreatic resection [109–114] (Table 1). Two studies have reported on the association between prehabilitation and post-operative outcomes in pancreatic cancer; Ausania et al. randomised patients to prehabilitation (*n* = 18) or standard of care (*n* = 22). Nakajima et al. compared a cohort of patients undergoing prehabilitation to historical patients. Neither study determined a significant difference in outcomes except for reduced rates of delayed gastric emptying in one and a shorter length of stay in the other. Several studies report an improvement in lean muscle mass prior to surgery and one reported improvement in quality of life in those undergoing prehabilitation [109–114]. Although showing promising results, the collective interpretation of these studies is difficult owing to poor standardisation of exercise and nutritional interventions and paucity of participants. Table 4: Pre-operative areas for potential gain.


**Table 3.** Summary of prehabilitation studies for pancreatic resection.

6MWT = 6 min walk test, STS = Sit to stand, GS = Gait speed, HRQOL = Health Related Quality of Life, LPA = Light physical activity.

**Table 4.** Pre-operative areas for potential gain.


PEI = Pancreatic exocrine insufficiency, PBD = Pre-operative biliary drainage, QoL = Quality of life, PERT = Pancreatic enzyme replacement therapy, PPI = Proton pump inhibitor, CGA = Comprehensive geriatric assessment.

#### **3. Peri-Operative Pathways**

#### *3.1. Enhanced Recovery after Surgery*

Enhanced recovery after surgery (ERAS) is a multimodal approach to a patient's perioperative journey which aims to facilitate early return to the preoperative state [115]. Broadly, it facilitates sustained recovery and reduces complications. A secondary benefit is frequently a reduced length of stay.

The value of ERAS pathways in pancreatic surgery has been recommended on the basis of high level evidence in domains such as avoidance of hypothermia, use of wound catheters compared to epidural analgesia (EDA), use of somatostatin analogues to reduce CR-POPF [116]. protocols for thromboprophylaxis and antimicrobials and interventions for preoperative nutrition for patients with severe weight loss [36]. As patient reported outcomes (PROs) have become of particular importance in pancreatic cancer due to the elderly cohort of patients, a recent addition has been patient-centred PROs into the ERAS pathway [117,118].

The degree of compliance with ERAS pathways is strongly associated with clinical outcome [119]. With numerous components, intensive monitoring of the pathway with regular audit and a dedicated specialist nurse input improves compliance. An evaluation into the feasibility of an ERAS pathway after PD has demonstrated over 70% compliance achieved within a multicentre cohort study and was associated with a significant reduction of overall complications and length of stay [120].

ERAS after pancreatic surgery have been associated with a reduction in mild complications (Clavien Dindo grade I-II) significant improvements in overall morbidity and length of stay without any increase in readmission [121–125]. Effect on pancreatic specific complications is difficult to interpret as studies have variably included PD and distal pancreatectomy and ISGPS definitions have not been consistently reported for delayed gastric emptying (DGE) and postoperative pancreatic fistula (POPF) However, a lower incidence of DGE has been observed with no effect on POPF rate [122,123,125].

Implementation of ERAS pathways has been hindered due to the economic impact associated with the necessary resources required, namely a specialist nurse, audit and data collection and patient information booklets [126]. However, these costs, have been offset by the reduction in postoperative complications and subsequent hospital length of stay [127].

Though the short-term benefits are evident, long term benefits have also been suggested by one study where a relationship between increased compliance to the ERAS pathway and survival benefit has been found [128].

#### *3.2. Reducing Complications from Surgery*

Surgery alone, is a poor treatment for pancreatic cancer. Only with associated receipt of chemotherapy do patients gain a good chance for cure. One major barrier to patients receiving adjuvant therapy is the occurrence of post-operative complications. Thus, strategies to reduce these are attractive not only to improve perioperative outcomes but improve the delivery of adjuvant therapy. Post-operative pancreatic fistula (POPF) is the most frequent and severe complication after surgery [129]. Presently there is no widely accepted approach to reducing rates of clinically relevant POPF but a national study of early detection and minimally invasive treatment of POPF aims to determine whether the severity of POPF can be reduced [130].

An individual patient's risk of POPF varies hugely. It is somewhat remarkable that POPF rates are not routinely adjusted to take this into account. Individual surgeons risk adjustment and CUSUM analysis is a way for surgeons to objectively assess their outcomes [131]. Such strategies could help inform surgeons of optimal techniques.

Strategies to improve outcomes after complex procedures such as pancreatoduodenectomy, evolve from a critical understanding of events and outcomes. Without critical analysis background noise, variations in practice and organisational differences can make this task insensitive. Determining the root causes of mortality after pancreatectomy demonstrates this well [129]. Only by conducting in depth analyses of processes and outcomes

can surgeons begin to understand fundamental reasons for failure. Solutions can then be developed which are designed to overcome problems which are prevalent at the local level. Solutions are likely to vary from centre to centre dependent upon variation in practice and outcome.

A further key improvement is to determine benchmarks of optimal outcomes. This strategy seeks to reduce the effect of variation in practice between centres by focussing on common factors between centres and avoiding outlier cases. In pancreatic cancer, this has allowed teams to compare outcomes to those of their peers after surgery for resectable cancer or with associated venous resection [132,133].

Taking the concept of assimilated gains, the Dutch are leading the way with a nationwide implementation of best practices based upon critical analysis of pathways and suboptimal outcomes with the PACAP-1 trial which seeks to improve outcomes and overall survival [134]. Table 5: Summary of Peri-operative areas for gain


**Table 5.** Summary of Peri-operative areas for gain.

ERAS = Enhanced Recovery after Surgery, QI = Quality Improvement, POPF = Post-Operative Pancreatic Fistula, LoS = Length of Stay, PERT = Pancreatic Exocrine Replacement Therapy.

#### **4. Post-Operative Pathways**

Post-operative strategies should be designed to enhance functional recovery, maximise uptake of adjuvant chemotherapy and delivering appropriate surveillance. Care should be continuous and multidisciplinary, with continued nutritional consideration (as outlined in the pre-operative section). Functional decline and lack of chemotherapy uptake is more pronounced in elderly.

#### *4.1. Adjuvant Therapy in Resectable Pancreatic Cancer*

The benefit for adjuvant therapy is without question. Alone surgery achieves cure in less than 10% of patients [3,135]. Currently, a multimodal approach is the standard of care where 6 months of mFOLFIRINOX based on the results of PRODIGE-24 RCT for those with a sufficient performance status, or a combination of gemcitabine and capecitabine based on ESPAC-4 RCT [136,137]. Patients not sufficiently fit for combination therapy may still benefit from gemcitabine. The role of adjuvant therapy following neoadjuvant therapy and resection, however, is less clear. A recent multi-centre international study demonstrated

that adjuvant chemotherapy following NAT, was only of benefit among patients with node positive disease [138].

Widespread variation in the use of adjuvant therapy is clear between and within countries. For example, studies demonstrate that 51% of patients receive adjuvant therapy in the USA, 54% in the Netherlands, 66% in Japan and 74% in Canada. Widespread variation within countries is also evident [139–142]. Within the Netherlands, rates of adjuvant therapy varied from 26 to 74% between health care providers. Sociodemographic variation explains some variation with deprived patients less likely to receive therapy [143]. Advanced age is a common factor associated with underuse of adjuvant therapy and will be considered separately below. However, such wide differences cannot be explained by demographic differences alone. Such data clearly points at systematic variation. Such variation is clearly undesirable and yet, despite the clear advantage of adjuvant therapy, there is little emphasis upon ensuring that patients chances of receiving therapy are optimized [144]. This low completion rate of the full therapeutic sequence may in part be explained by the reticence to initiate adjuvant chemotherapy following postoperative complications or poor performance status. However, ESPAC-3 has addressed this issue of time to initiate and optimal duration of chemotherapy [145]. Within this study, 68% of patients completed all six cycles of chemotherapy and for these patients, overall survival was significantly favoured. In those who did complete therapy, there was no difference in survival when comparing those who started earlier than eight weeks compared to those who started between 8 and 12 weeks, therefore time to initiation of chemotherapy was an important prognostic factor in favour of later treatment. Thus, this study concluded that completion of chemotherapy rather than early initiation was more important for survival.

Centralised cancer surgery is widely practiced. However, centralised chemotherapy is not standard practice. There is evidence that concentrating adjuvant therapy to a dedicated regional service increases the proportion of patients that receive therapy [146]. Overcoming nihilistic views and patients fear of therapy are important strategies as many patients choose not to pursue chemotherapy and clinicians not referring patients for chemotherapy is a mindset in great need of change [147].

A further variable that must be considered is not simply whether a patient receives adjuvant therapy but that efforts must focus upon an individualised approach where patients receive a regime that is as strong as can be tolerated for that individual. Incremental gains of multiagent therapy are seen over single agent gemcitabine [3,135,137]. Yet, there is very little data upon strategies to optimise not just the delivery of adjuvant therapy but also the regimen that is delivered. Some evidence supports centralised care to deliver more multiagent therapy [147].

Liquid biopsies are a novel way to diagnose cancer at an early stage, aid in prognostic evaluation [148], determine targets for therapy [149] and to evaluate cancer recurrence after treatment. The technique involves determining cancer DNA, vesicles and tumour cells in circulating blood [150,151]. Nomograms can be used to stratify patient risk for selection to treatment and may influence the choice of NAT, surgery or nature of adjuvant chemotherapy [152–156].

#### *4.2. Benefits of and Access to Chemotherapy in the Elderly Population*

After resection of pancreatic cancer, disease free survival (DFS) and disease specific survival (DSS) appear similar between elderly and younger patients, but overall survival (OS) is shorter, a possible reflection that elderly patients are less likely to receive adjuvant chemotherapy [86,157–159]. Elderly patients have just as much benefit from adjuvant therapy as younger patients and when chemotherapy use is stratified between young and older patients overall survival is the same [160]. The CONKO-001 trial had no upper age limit and specifically demonstrated that those over 65 show a similar improvement in OS and DFS as have numerous other studies evaluating adjuvant chemotherapy for pancreatic cancer [137,159,161–163]. This benefit is maintained when considering more aggressive regimens in the elderly such as FOLFIRINOX or neo-adjuvant chemotherapy [164–166].

Consequently, strategies to increase the use of chemotherapy in the elderly, such as centralisation of oncology services and measures to address frailty, yield major benefits [146,167]. The comprehensive geriatric assessment (CGA) is recommended by the Society for International Oncology in Geriatrics as a useful decision making tool for older people with malignancy [168]. There are several studies suggesting that the CGA can be useful in predicting functional decline, toxicity and overall survival [169,170] The CGA is lengthy and time consuming, a more practical approach, as recommended by the National Comprehensive Cancer Network, the European Organisation for Research and Treatment of Cancer and the International Society of Geriatric Oncology is to use a short frailty screening tool (such as the Fried score, the Clinical Frailty Scale or the Geriatric 8) to identify those who require a full screening [171–177].

Like surgery, the benefits of chemotherapy are not lessened by age and therefore older patients with appropriate performance status should not be denied access. Treatment decisions should be made in a multidisciplinary setting and include the use of a frailty screening tool and the input of a geriatrician to avoid discrepancies in treatment based on chronological age [95,96].

Since there are no published accepted standards for the proportion of patients that receive adjuvant therapy after pancreatic cancer surgery there is lack of clarity about what is acceptable practice. Thus, local audit and benchmarking one's practice against peers is an essential step in the standardisation of practices. Quality improvement programs could further enhance delivery of adjuvant chemotherapy if they were to target those patients most at risk of not receiving therapy, i.e., to overcome barriers presented by age, frailty and good functional recovery after surgery.

#### *4.3. Surveillance after Resection of Pancreatic Cancer*

Surveillance after resection of other cancer types is routine and evidence based. The notion is simple, that early detection of recurrence is more likely to identify disease at an earlier asymptomatic phase when patients are more likely to have preserved performance status and ultimately more likely to receive treatment. Most guidelines, including those from the European Society for Medical Oncology (ESMO) and International Association of Pancreatology/European Pancreatic Club (IAP/EPC), do not recommend routine surveillance after pancreatic cancer resection due to the poor prognosis and limited treatment options available for recurrence [178,179]. The nature of surveillance protocols typically involve CT scans at 3 or 6 month intervals supplemented by CA19-9 analysis. The optimal interval and impact of surveillance programs remain to be seen. It may be that artificial intelligence may help refine and improve surveillance programs; there is currently much interest in AI as a tool to facilitate the diagnosis of breast cancer within screening programs.

In the setting of recurrent pancreatic cancer there are an increasing number of treatment strategies. Despite this, many patients fail to receive palliative therapy. However, an individualised approach to treating recurrent pancreatic cancer is logical. Around half of all patients recur with local only disease [180]. Whilst systemic therapy seeks to control occult metastatic disease local therapies in the form of high intensity radiotherapy or ablation can be delivered [181]. Given that many patients are elderly, frail or may have had poor experience with systemic therapy local therapy can be delivered with more less disruption and more satisfaction for the patient. Targeted therapy of oligometastatic disease also offers alternative options for a limited number of patients. Furthermore, surveillance is desired by patients and clinicians alike [182,183]. Table 6: Areas for Post-operative gain.


**Table 6.** Areas for Post-operative gain.

QI = Quality improvement.

#### **5. Novel Areas for Review**

Though this body of work focusses upon areas where benefit can be clearly derived using existing evidence base there are areas with an emerging evidence base where benefit may be derived if practices are adopted.

Pancreatic surgery is associated with a high rate of cancer recurrence. Improving the staging pathways and novel approaches to treatment of the surgical margin are ways to improve outcomes.

Early cancer recurrence, even within ninety days of surgery, has to be considered a failure of staging [129]. The addition of routine Positron Emission Topography (PET) or Magnetic resonance imaging (MRI) scanning can help detect occult cancer. PET scanning, within a randomised trial, upstaged 20% of patients, preventing futile surgery [184]. However, it is important to note that the routine use of PET-CT is still debatable as it cause delays to resection and the potentially false positive results that may erroneously prevent surgery [185,186]. MRI can upstage 10–24% of patients with occult liver metastases where diffusion weight images can identify lesions under 5 mm, being more sensitive than computerised tomography (CT) or PET-CT [187–189]. These tests need to be considered carefully, however, as false positive results can be caused by common scenarios such as abscesses secondary to cholangitis and can delay treatment.

Failure of surgery to clear the margin is an Achilles heel of pancreatoduodenectomy. Local recurrence is very high after surgery, positive margins are associated with reduced survival and therefore it is necessary to consider how outcomes can be improved [190,191]. Neoadjuvant therapy (NAT) is associated with a higher rate of R0 resections, though it is important to consider results on an intention to treat basis [192,193]. There is much interest in NAT for resectable pancreatic cancer. Ongoing clinical trials will help determine the evidence base for this treatment. There are many retrospective cohort studies which are at risk of various bias including selection and survivor bias and until well conducted trial data is available it is not possible to make a clear statement [194]. Intraoperative frozen section has been used to identify positive pancreatic transection margins and the role of extending resections has been explored in numerous studies, but systematic review fails to observe a benefit to this practice [193]. Similarly, extending lymphadenectomy is not associated with greater survival but with morbidity [195,196]. The standard approach to

pancreatoduodenectomy leaves perineural tissue around the superior mesenteric artery. It is this margin which is most frequently positive on histologic analysis. An artery first approach has a theoretical advantage that it can clear periadventitial tissues but clinical studies fail to demonstrate benefit in terms of improving R0 margin status [197,198]. Intraoperative radiotherapy has been long considered as an adjunct to improving margin status but no high-level evidence yet exists [199]. The addition of stereotactic ablative radiotherapy within a NAT program does not appear to improve R0 rates [200]. Irreversible electroporation has gained interest, predominantly surrounding control of locally advanced pancreatic cancer though the technology could be applied to improve margin control at resectable cancer. However, concerns over safety and a lack of high level evidence remain [201]. There is thus much work to be done in this area and focussing efforts on the problem of the surgical margin could reduce local recurrence.

Peritoneal metastases are less common than local recurrent or liver metastases but when they occur are associated with very poor outcomes. Positive intraoperative peritoneal cytology correlates with poor oncologic outcomes, in the absence of visible metastatic disease [202]. Among patients with cytology positive, or even macroscopic peritoneal disease, intraperitoneal chemotherapy can control ascites and some patients can go on to achieve surgical resection [203,204]. Such experience is limited to Japan. The remarkable outcomes indicate a potential role for intraperitoneal therapy, challenge current beliefs about pancreatic cancer biology and prognosis and merit a wider review.

Personalised medicine in pancreatic cancer care is most strongly associated with defining an individual's cancer genetics. Large multinational, multicentre organisations/trials are ongoing such as precisionpanc (https://precisionpanc.org, accessed on 10 January 2021). There have, however, been many negative trials of genetic targets among patients with metastatic pancreatic cancer which include MEK inhibitors, IGFR inhibitors, mTOR inhibitors, TRK inhibitors, NOTCH inhibitors, TGF-β inhibitors, immunotherapy or vaccine therapy [205]. Established genetic targets in resectable pancreatic cancer are thus largely unclear; the most actionable mutation is BRCA1-2 which is identified in approximately 5% of patients [206]. Patients with these tumours are sensitive to platinum agents and thus will be treated by inclusion of oxaliplatin in therapy regimens [207]. This target has also been exploited by the PARP inhibitor Olaparib [208].

#### **6. Conclusions**

This review sought to elucidate key areas of variation, undertreatment and pathway changes where improvements can be realised with little effort. Novel therapeutic options will present themselves in the future, but it would be remiss of any team caring for this cohort of patients to inadequately utilise current evidence and implement optimal treatment pathways. There is a disconnect between funding for research to establish novel treatments far outstripping funding to implement best care at the level of the health care provider/organisation. It is essential that surgeons understand that surgery is just one part of a complex pathway and that they are ideally placed to act as change agents to optimise broader pathway improvements. Some changes can be clearly applied to the majority, if not all patients, such as PERT among those undergoing pancreatoduodenectomy. Other interventions can never be applied to all patients. There will always be some jaundiced patients with resectable cancer that undergo PBD or some patients who never receive adjuvant therapy. Locally, nationally or internationally accepted benchmarks are required to understand what is achievable and to help teams identify areas of poor performance. Collaborative multicentre, multinational studies are an essential part of assessing and improving patient care in the 21st century and teams are encouraged to develop and take part in these ventures. Through the aggregation of marginal gains, our patients can realise better outcomes and experience in the near future.

**Author Contributions:** K.J.R.: Conception and design, Supervision, Writing—Review and editing; R.P. and S.P.-B.; Research and writing—original draft preparation, in equal parts. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


*Article*

## **Predictive Features of Malignancy in Branch Duct Type Intraductal Papillary Mucinous Neoplasm of the Pancreas: A Meta-Analysis**

**Wooil Kwon** † **, Youngmin Han** † **, Yoonhyeong Byun , Jae Seung Kang , Yoo Jin Choi, Hongbeom Kim and Jin-Young Jang \***

Department of Surgery and Cancer Research Institute, Seoul National University College of Medicine, Seoul 03080, Korea; willdoc@snu.ac.kr (W.K.); views@snu.ac.kr (Y.H.); yoonhyeong@snu.ac.kr (Y.B.); 74398@snuh.org (J.S.K.); 74401@snuh.org (Y.J.C.); surgeonkhb@snu.ac.kr (H.K.)

**\*** Correspondence: jangjy4@snu.ac.kr; Tel.: +82-2-2072-2914; Fax: +82-2-766-3975

† These authors contributed equally to this paper as first authors.

Received: 31 July 2020; Accepted: 11 September 2020; Published: 14 September 2020

**Simple Summary:** Currently, there are several guidelines that are widely used to establish the treatment strategy for branch duct type intraductal papillary mucinous neoplasms. Although there are some common grounds, there are discrepancies on which features they adopt, how much each feature is weighted, and how the features are combined. Furthermore, some of the features are based on lower level evidences or expert opinions. The aim of this meta-analysis was to investigate important clinical, radiological, and biochemical risk factors for malignancy and their impact as predictors. This study found symptom, size, cyst wall thickening, presence of mural nodule, change in main pancreatic duct caliber, lymphadenopathy, CA 19-9, and CEA as risk factors. Lymphadenopathy (odd ratio [OR]: 8.55), abrupt caliber change (OR: 7.41), and mural nodule (OR: 4.10) had the highest odd ratios. We expect the higher level evidences of this study to help shape better guidelines and reduce discrepancies among future guidelines.

**Abstract:** The current guidelines on branch duct type intraductal papillary mucinous neoplasm (BD-IPMN) recommend various predictive features of malignancy as well as different treatment strategies. This study aimed to identify the risk factors for malignancy with higher level of evidence. A meta-analysis was performed on 40 literatures published between 2000 and 2019. These literatures included 6301 patients with pathologically proven IPMN. Malignancy was defined as high-grade dysplasia and invasive carcinoma. It was significantly associated with symptoms (odds ratio [OR] 1.35, confidence interval [CI] 1.01–1.79), size ≥ 3 cm (OR 1.90, CI 1.51–2.40), cystic wall thickening (OR 2.53, CI 1.50–4.27), mural nodule (OR 4.10, CI 3.38–4.97), main pancreatic duct dilatation (OR 2.98, CI 2.11–4.21), abrupt caliber change of the pancreatic duct (OR 7.41, CI 2.49–22.06), lymphadenopathy (OR 8.55, CI 3.25–22.51), elevated carbohydrate antigen 19-9 (OR 4.01, CI 2.55–6.28), and elevated carcinoembryonic antigen (OR 2.04, CI 1.60–2.61). Multilocular cysts and multiple cysts did not show a significant association with malignancy. This study examined the clinical, radiological, and biochemical features of BD-IPMN, often used as malignancy predictors according to the widely used guidelines. The results confirmed that all the features currently being used are valid.

**Keywords:** branch duct intraductal papillary mucinous neoplasm; risk factor; malignancy; meta-analysis

#### **1. Introduction 1. Introduction**

Branch duct type intraductal papillary mucinous neoplasm (BD-IPMN) is a well-known premalignant lesion of the pancreas. The prevalence of BD-IPMN-associated malignancy is reportedly over 24% [1]. Nearly four decades have passed since the first report of IPMN by Ohashi et al. [2], but our understanding of IPMN is still limited. Particularly, the ability to predict malignancy and set an appropriate treatment plan is far from satisfactory. Given that pancreatic cancer is the fourth leading cause of cancer mortality [3], the clinical implications of this shortcoming are grave. To make matters worse, the incidence of IPMN is on a steady rise, as incidental detections are increasing due to better access to heath check-ups and increased use of cross-sectional imaging studies [4,5]. The current situation poses a great challenge for pancreatic surgeons and physicians. Branch duct type intraductal papillary mucinous neoplasm (BD-IPMN) is a well-known premalignant lesion of the pancreas. The prevalence of BD-IPMN-associated malignancy is reportedly over 24% [1]. Nearly four decades have passed since the first report of IPMN by Ohashi et al. [2], but our understanding of IPMN is still limited. Particularly, the ability to predict malignancy and set an appropriate treatment plan is far from satisfactory. Given that pancreatic cancer is the fourth leading cause of cancer mortality [3], the clinical implications of this shortcoming are grave. To make matters worse, the incidence of IPMN is on a steady rise, as incidental detections are increasing due to better access to heath check-ups and increased use of cross-sectional imaging studies [4,5]. The current situation poses a great challenge for pancreatic surgeons and physicians.

*Cancers* **2020**, *12*, x 2 of 18

Many investigations have been conducted, results have been produced, and the endeavor continues. Currently, there are several management guidelines for IPMN. Among them, the most frequently referenced are those by the American Gastroenterological Association (AGA) [6], European Study Group on Cystic Tumours of the Pancreas [7,8], and International Association of Pancreatology (IAP) [1,9,10]. Although all these guidelines have some commonalities, they do differ with respect to certain surgical treatment indications and surveillance strategies. Another issue is that few of these guidelines cite studies with lower levels of evidences, while others cite experts' opinions. Many investigations have been conducted, results have been produced, and the endeavor continues. Currently, there are several management guidelines for IPMN. Among them, the most frequently referenced are those by the American Gastroenterological Association (AGA) [6], European Study Group on Cystic Tumours of the Pancreas [7,8], and International Association of Pancreatology (IAP) [1,9,10]. Although all these guidelines have some commonalities, they do differ with respect to certain surgical treatment indications and surveillance strategies. Another issue is that few of these guidelines cite studies with lower levels of evidences, while others cite experts' opinions.

The first step in producing high-quality treatment guidelines for BD-IPMN is to clarify the risk factors for malignancy. Therefore, a meta-analysis was performed to identify the clinically important risk factors for malignancy and their impact. This study investigated the comprehensive factors including clinical, radiological, and biochemical factors that could be acquired preoperatively. The first step in producing high-quality treatment guidelines for BD-IPMN is to clarify the risk factors for malignancy. Therefore, a meta-analysis was performed to identify the clinically important risk factors for malignancy and their impact. This study investigated the comprehensive factors including clinical, radiological, and biochemical factors that could be acquired preoperatively.

#### **2. Results 2. Results**

#### *2.1. Search Results 2.1. Search Results*

The search process is described in Figure 1. A thorough literature search on MEDLINE identified 472 publications that were potentially relevant to this study. A total of, 412 studies were excluded after screening. Of the remaining 60 publications, 17 were excluded after detailed review due to insufficient data regarding worrisome features/high-risk stigmata, absence of pathological data, insufficient sample size, or overlap with another study. When an overlapping study cohort was found, the larger sample study was chosen. If there were results regarding worrisome features/high-risk stigmata in a smaller overlapping study that was not addressed in the larger one, it was still included. Finally, 40 publications were included in the analysis [11–50]. The publication bias was assessed visually by inspecting the funnel plot for asymmetry. The search process is described in Figure 1. A thorough literature search on MEDLINE identified 472 publications that were potentially relevant to this study. A total of, 412 studies were excluded after screening. Of the remaining 60 publications, 17 were excluded after detailed review due to insufficient data regarding worrisome features/high-risk stigmata, absence of pathological data, insufficient sample size, or overlap with another study. When an overlapping study cohort was found, the larger sample study was chosen. If there were results regarding worrisome features/highrisk stigmata in a smaller overlapping study that was not addressed in the larger one, it was still included. Finally, 40 publications were included in the analysis [11–50]. The publication bias was assessed visually by inspecting the funnel plot for asymmetry.

**Figure 1. Figure 1.** A flow diagram of the inclusion criteria of studies eligible for meta-analysis. A flow diagram of the inclusion criteria of studies eligible for meta-analysis.

#### *2.2. Characteristics of Included Studies*

The characteristics of the included publications are described in Table 1. This study included 40 publications with 6301 patients diagnosed with IPMN, of which histological data of 4241 patients diagnosed with benign IPMN and 2060 with malignant IPMN were identified. In all studies, malignant IPMN was defined as invasive carcinoma and high-grade dysplasia. Terms such as invasive cancer, intraductal papillary mucinous carcinoma (IPMC), and invasive IPMN were considered equivalent to invasive carcinoma. Non-invasive carcinoma, carcinoma in situ, and IPMC in situ were considered equivalent to high-grade dysplasia.

#### *2.3. Clinical Symptoms*

Data regarding symptoms were extractable in 15 studies [12,15,19,20,23,25,27,31,33,35,36,38,40–42]. In these studies, 840 patients (54.8%) presented symptoms, and malignancy was reported in 28.6% of patients with symptoms and 27.4% without symptoms. The odds ratio (OR) of having symptoms was 1.35 (95% confidence interval [CI] 1.01–1.79, *p* = 0.040) (Table 2, Figure 2a).

#### *2.4. Characteristics of Cyst*

Data regarding cyst size were obtained from 22 studies with 4446 patients [12,15–17,20–22,24–26, 30–33,36–38,40,41,43,48,49], and the risk of malignancy was examined for a reference size of 3 cm. The malignancy rate in cysts ≥ 3 cm and < 3 cm in size was 38.7% and 25.7%, respectively. Cysts of size ≥ 3 cm significantly increased the risk of malignancy with an OR of 1.90 (95% CI 1.51–2.40, *p* < 0.001) (Table 2, Figure 2b).

Data regarding cystic wall thickening was extracted from nine studies with 689 patients [13,14,17,18,34,37,38,40,43], and wall thickening was found in 15.2% of the cases. Moreover, 51.4% of the patients with wall thickening reported malignancy as compared to 23.6% of those without wall thickening. Wall thickening was significantly associated with malignancy with OR of 2.53 (95% CI 1.50–4.27, *p* < 0.001) (Table 2, Figure 2c).

Multilocularity and multiplicity was analyzed in seven [17,20,21,34,37,40,47] and eight studies [12,15,17,21,35,37,40,47], respectively. Malignancy rate of multilocular and unilocular cysts was 27.0% and 22.2%, respectively. Furthermore, the malignancy rate of single and multiple cysts was 26.6% and 24.0%, respectively. Notably, neither of the features was associated with an increased risk of malignancy (multilocularity: OR 0.92, 95% CI 0.63–1.35, *p* = 0.680; multiplicity: OR 0.76, 95% CI 0.55–1.04, *p* = 0.090) (Table 2, Figure 2d,e).




N/A, not available; IPMN, intraductal papillary mucinous neoplasm; BD, branch duct; MD, main duct; CT. computed tomography; MRI, magnetic resonance image; MRCP, magnetic resonance cholangiopancreatography; ERCP, endoscopic retrograde cholangiopancreatography; US, ultrasonography; CEUS, contrast-enhanced ultrasonography; EUS, endoscopic ultrasonography; CE-EUS, contrast-enhanced endoscopic ultrasonography; FNA, fine needle aspiration. \* in median.

*Cancers* **2020**, *12*, 2618



*2.2. Characteristics of Included Studies* 

*2.3. Clinical Symptoms* 

*2.4. Characteristics of Cyst* 

0.001) (Table 2, Figure 2b).

1.50–4.27, *p* < 0.001) (Table 2, Figure 2c).

1.04, *p* = 0.090) (Table 2, Figures 2d,e)

in situ were considered equivalent to high-grade dysplasia.


The characteristics of the included publications are described in Table 1. This study included 40 publications with 6301 patients diagnosed with IPMN, of which histological data of 4241 patients diagnosed with benign IPMN and 2060 with malignant IPMN were identified. In all studies, malignant IPMN was defined as invasive carcinoma and high-grade dysplasia. Terms such as invasive cancer, intraductal papillary mucinous carcinoma (IPMC), and invasive IPMN were considered equivalent to invasive carcinoma. Non-invasive carcinoma, carcinoma in situ, and IPMC

Data regarding symptoms were extractable in 15 studies [12,15,19,20,23,25,27,31,33,35,36,38,40– 42]. In these studies, 840 patients (54.8%) presented symptoms, and malignancy was reported in 28.6% of patients with symptoms and 27.4% without symptoms. The odds ratio (OR) of having

Data regarding cyst size were obtained from 22 studies with 4446 patients [12,15–17,20–22,24– 26,30–33,36–38,40,41,43,48,49], and the risk of malignancy was examined for a reference size of 3 cm. The malignancy rate in cysts ≥ 3 cm and < 3 cm in size was 38.7% and 25.7%, respectively. Cysts of size ≥ 3 cm significantly increased the risk of malignancy with an OR of 1.90 (95% CI 1.51–2.40, *p* <

Data regarding cystic wall thickening was extracted from nine studies with 689 patients [13,14,17,18,34,37,38,40,43], and wall thickening was found in 15.2% of the cases. Moreover, 51.4% of the patients with wall thickening reported malignancy as compared to 23.6% of those without wall thickening. Wall thickening was significantly associated with malignancy with OR of 2.53 (95% CI

Multilocularity and multiplicity was analyzed in seven [17,20,21,34,37,40,47] and eight studies [12,15,17,21,35,37,40,47], respectively. Malignancy rate of multilocular and unilocular cysts was 27.0% and 22.2%, respectively. Furthermore, the malignancy rate of single and multiple cysts was 26.6%

symptoms was 1.35 (95% confidence interval [CI] 1.01–1.79, *p* = 0.040) (Table 2, Figure 2a).



**Figure 2.** *Cont.*


*Cancers* **2020**, *12*, x 5 of 18

**Figure 2.** Forest plot showing the incidence of malignancy in BD-IPMN stratified by symptoms, characteristics of the cyst, and presence of mural nodule. (**a**) symptom. (**b**) cyst size. (**c**) cyst wall thickening. (**d**) multilocular cyst. (**e**) multiple cyst. and (**f**) mural nodule. **Figure 2.** Forest plot showing the incidence of malignancy in BD-IPMN stratified by symptoms, characteristics of the cyst, and presence of mural nodule. (**a**) symptom. (**b**) cyst size. (**c**) cyst wall thickening. (**d**) multilocular cyst. (**e**) multiple cyst. and (**f**) mural nodule.

#### *2.5. Mural Nodule*

Mural nodule was the most frequently investigated parameter observed in 25 studies and cohort of 4495 patients [12,13,16–26,30–33,35,36,38,40,42,43,46,47]. The prevalence of mural nodule in BD-IPMN was 35.8%, and the pooled malignancy rate was 31.9%. The malignancy rate was 52.5% in the presence of mural nodule and 20.4% in its absence. The presence of mural nodule resulted in a four-fold increase in the malignancy risk. The pooled OR was 4.10 (95% CI 3.38–4.97, *p* < 0.001) (Table 2, Figure 2f).

#### *2.6. Changes in Main Pancreatic Duct*

Several studies examined the size of the main pancreatic duct, but they all had different cut-off values. The reference size was 5 mm in eight studies [22,24–26,37,38,40,43], 6 mm in five studies [16,30–32,34], and 7 mm in two studies [12,36]. For pancreatic ducts of size 5 mm, the OR was 2.85 (95% CI 1.90–4.26, *p* < 0.001), and a malignancy rate of 46.5% for ducts > 5 mm. The ORs for main pancreatic ducts >6 and 7 mm were 3.86 (95% CI 1.63–9.11, *p* = 0.002) and 2.69 (95% CI 0.42–17.16, *p* = 0.29), respectively. Overall, the OR for dilatation of the main pancreatic duct was 2.98 (95% CI 2.11–4.21, *p* < 0.001) (Table 2, Figure 3a).

*2.5. Mural Nodule* 

Figure 2f).

*2.6. Changes in Main Pancreatic Duct* 

4.21, *p* < 0.001) (Table 2, Figure 3a).

with OR of 7.41 (95% CI 2.49–22.06, *p* < 0.001) (Table 2, Figure 3b).


Mural nodule was the most frequently investigated parameter observed in 25 studies and cohort of 4495 patients [12,13,16–26,30–33,35,36,38,40,42,43,46,47]. The prevalence of mural nodule in BD-IPMN was 35.8%, and the pooled malignancy rate was 31.9%. The malignancy rate was 52.5% in the presence of mural nodule and 20.4% in its absence. The presence of mural nodule resulted in a fourfold increase in the malignancy risk. The pooled OR was 4.10 (95% CI 3.38–4.97, *p* < 0.001) (Table 2,

Several studies examined the size of the main pancreatic duct, but they all had different cut-off values. The reference size was 5 mM in eight studies [22,24–26,37,38,40,43], 6 mM in five studies [16,30–32,34], and 7 mM in two studies [12,36]. For pancreatic ducts of size 5 mM, the OR was 2.85 (95% CI 1.90–4.26, *p* < 0.001), and a malignancy rate of 46.5% for ducts > 5 mM. The ORs for main pancreatic ducts > 6 and 7 mM were 3.86 (95% CI 1.63–9.11, *p* = 0.002) and 2.69 (95% CI 0.42–17.16, *p* = 0.29), respectively. Overall, the OR for dilatation of the main pancreatic duct was 2.98 (95% CI 2.11–

Four studies examined the caliber change in the pancreatic duct in 467 patients [14,37,38,46].


**Figure 3.** Forest plot demonstrates the incidence of malignancy in BD-IPMN in relation to change in main pancreatic duct. Forest plot stratified (**a**) by the diameter of pancreatic duct and (**b**) by abrupt caliber change. main pancreatic duct. Forest plot stratified (**a**) by the diameter of pancreatic duct and (**b**) by abrupt caliber change.

*2.7. Lymphadenopathy*  Four studies examining lymphadenopathy had a pooled cohort of 390 patients [14,17,21,43]. The prevalence of lymphadenopathy was 6.2%. The malignancy rate in these patients was 58.3% as Four studies examined the caliber change in the pancreatic duct in 467 patients [14,37,38,46]. Among 34 patients with an abrupt change in caliber, 18 patients (52.9%) had malignant BD-IPMN with OR of 7.41 (95% CI 2.49–22.06, *p* < 0.001) (Table 2, Figure 3b).

#### compared to 15.3% in those without lymphadenopathy. The OR for lymphadenopathy was the *2.7. Lymphadenopathy*

of lymphadenopathy.

*2.8. Biochemical Markers* 

highest among all parameters at 8.55 (95% CI 3.25–22.51, *p* < 0.001) (Table 2, Figure 4) CEA was 2.04 (95% CI 1.60–2.61, *p* < 0.001) (Table 2, Figure 5b). **Figure 3.** Forest plot demonstrates the incidence of malignancy in BD-IPMN in relation to change in Four studies examining lymphadenopathy had a pooled cohort of 390 patients [14,17,21,43]. The prevalence of lymphadenopathy was 6.2%. The malignancy rate in these patients was 58.3% as compared to 15.3% in those without lymphadenopathy. The OR for lymphadenopathy was the highest among all parameters at 8.55 (95% CI 3.25–22.51, *p* < 0.001) (Table 2, Figure 4).

**Figure 4.** Forest plot demonstrates the incidence of malignancy in BD-IPMN stratified by the presence

Carbohydrate antigen (CA) 19-9 with a cut-off level of 37 U/mL was examined in eight studies [15,21,23–26,36,40]. Among 3279 pooled patients, 477 patients (14.5%) had elevated CA 19-9 levels, of which 61.8% had malignant BD-IPMN, whereas only 27.8% of the normal CA 19-9 patients showed

There were four studies [21,23–25] with pooled cohort of 2405 patients that reported the presence of carcinoembryonic antigen (CEA) with a cut-off level of 5 ng/mL. The malignancy rate among patients with elevated CEA and normal CEA was 53.5% and 35.7%, respectively. The OR for elevated

malignancy. The OR was 4.01 (95% CI 2.55–6.28, *p* < 0.001) (Table 2 and Figure 5a).

caliber change.

caliber change.

*2.7. Lymphadenopathy* 

*2.7. Lymphadenopathy* 


highest among all parameters at 8.55 (95% CI 3.25–22.51, *p* < 0.001) (Table 2, Figure 4)

highest among all parameters at 8.55 (95% CI 3.25–22.51, *p* < 0.001) (Table 2, Figure 4)

**Figure 3.** Forest plot demonstrates the incidence of malignancy in BD-IPMN in relation to change in main pancreatic duct. Forest plot stratified (**a**) by the diameter of pancreatic duct and (**b**) by abrupt

**Figure 3.** Forest plot demonstrates the incidence of malignancy in BD-IPMN in relation to change in main pancreatic duct. Forest plot stratified (**a**) by the diameter of pancreatic duct and (**b**) by abrupt

*Cancers* **2020**, *12*, x 10 of 18

Four studies examining lymphadenopathy had a pooled cohort of 390 patients [14,17,21,43]. The prevalence of lymphadenopathy was 6.2%. The malignancy rate in these patients was 58.3% as

Four studies examining lymphadenopathy had a pooled cohort of 390 patients [14,17,21,43]. The prevalence of lymphadenopathy was 6.2%. The malignancy rate in these patients was 58.3% as compared to 15.3% in those without lymphadenopathy. The OR for lymphadenopathy was the

**Figure 4.** Forest plot demonstrates the incidence of malignancy in BD-IPMN stratified by the presence of lymphadenopathy. **Figure 4.** Forest plot demonstrates the incidence of malignancy in BD-IPMN stratified by the presence of lymphadenopathy. Carbohydrate antigen (CA) 19-9 with a cut-off level of 37 U/mL was examined in eight studies [15,21,23–26,36,40]. Among 3279 pooled patients, 477 patients (14.5%) had elevated CA 19-9 levels, of

#### *2.8. Biochemical Markers 2.8. Biochemical Markers* which 61.8% had malignant BD-IPMN, whereas only 27.8% of the normal CA 19-9 patients showed malignancy. The OR was 4.01 (95% CI 2.55–6.28, *p* < 0.001) (Table 2 and Figure 5a).

Carbohydrate antigen (CA) 19-9 with a cut-off level of 37 U/mL was examined in eight studies [15,21,23–26,36,40]. Among 3279 pooled patients, 477 patients (14.5%) had elevated CA 19-9 levels, of which 61.8% had malignant BD-IPMN, whereas only 27.8% of the normal CA 19-9 patients showed malignancy. The OR was 4.01 (95% CI 2.55–6.28, *p* < 0.001) (Table 2 and Figure 5a). Carbohydrate antigen (CA) 19-9 with a cut-off level of 37 U/mL was examined in eight studies [15,21,23–26,36,40]. Among 3279 pooled patients, 477 patients (14.5%) had elevated CA 19-9 levels, of which 61.8% had malignant BD-IPMN, whereas only 27.8% of the normal CA 19-9 patients showed malignancy. The OR was 4.01 (95% CI 2.55–6.28, *p* < 0.001) (Table 2 and Figure 5a). There were four studies [21,23–25] with pooled cohort of 2405 patients that reported the presence of carcinoembryonic antigen (CEA) with a cut-off level of 5 ng/mL. The malignancy rate among patients with elevated CEA and normal CEA was 53.5% and 35.7%, respectively. The OR for elevated CEA was 2.04 (95% CI 1.60–2.61, *p* < 0.001) (Table 2, Figure 5b).



**Figure 5.** Forest plot demonstrates the incidence of malignancy in BD-IPMN in relation to biochemical markers. (**a**) carbohydrate antigen 19-9 and (**b**) carcinoembryonic antigen. **Figure 5.** Forest plot demonstrates the incidence of malignancy in BD-IPMN in relation to biochemical markers. (**a**) carbohydrate antigen 19-9 and (**b**) carcinoembryonic antigen.

**3. Discussion**  This study revealed that the parameters of symptoms, size, cystic wall thickening, presence of mural nodule, change in main pancreatic duct caliber, lymphadenopathy, CA 19-9, and CEA were There were four studies [21,23–25] with pooled cohort of 2405 patients that reported the presence of carcinoembryonic antigen (CEA) with a cut-off level of 5 ng/mL. The malignancy rate among patients with elevated CEA and normal CEA was 53.5% and 35.7%, respectively. The OR for elevated CEA was 2.04 (95% CI 1.60–2.61, *p* < 0.001) (Table 2, Figure 5b).

the predictive features of malignancy in BD-IPMN. On the other hand, multilocularity of cyst and

#### multiple cysts were not malignancy predictors. **3. Discussion**

The findings are in accordance with most of the widely used guidelines. The AGA guideline utilizes size, dilated main pancreatic duct, solid component, and positive cytology to determine the treatment strategy [6]. Reference size ≥ 3 cm, dilated main pancreatic duct, and associated solid component were considered risk factors, and presence of at least two of these would warrant endoscopic ultrasound-fine needle aspiration (EUS-FNA). In the case of positive cytology or presence This study revealed that the parameters of symptoms, size, cystic wall thickening, presence of mural nodule, change in main pancreatic duct caliber, lymphadenopathy, CA 19-9, and CEA were the predictive features of malignancy in BD-IPMN. On the other hand, multilocularity of cyst and multiple cysts were not malignancy predictors.

(< 5 mM) are relative indicators wherein healthy patients may opt for surgery.

Unlike the conservative AGA guidelines, the European study group proposes a more aggressive approach in BD-IPMN patients [7,8]. Presence of jaundice, positive cytology, enhancing mural nodule (≥ 5 mM), solid mass, and main pancreatic duct ≥10 mM are absolute indicators for surgery. Growth rate ≥ 5 mM/year, elevated serum CA 19-9 level, main pancreatic duct dilatation between 5–9.9 mM, cyst diameter ≥ 40 mM, new onset diabetes mellitus, acute pancreatitis, and enhancing mural nodule

The IAP guidelines stratify the features into high-risk stigmata and worrisome features. The high-risk stigmata and worrisome features warrant surgery and EUS, respectively. High-risk stigmata include obstructive jaundice in a patient with cystic lesion of the head of the pancreas, enhancing mural nodule ≥5 mM, and main pancreatic duct ≥ 10 mM. Worrisome features include cyst ≥3 cm, enhancing mural nodule <5 mM, thickened/enhancing cyst walls, main duct size 5–9 mM, abrupt change in caliber of pancreatic duct with distal pancreatic atrophy, lymphadenopathy, increased CA 19-9 serum level, and cystic growth rate ≥5 mM/2 years. The features used by the IAP and European study group are similar. However, the IAP guidelines are slightly more conservative,

In this study, all the features were included to validate those featuring in various guidelines. In addition, other features such as locularity, multiplicity, and CEA serum level were explored. The parameter of symptoms showed a significant association with malignancy. However, the symptoms could be heterogenous and often vague. They consisted of one or combinations of clinical findings such as abdominal pain, weight loss, pancreatitis, and jaundice. Therefore, it is difficult to define what symptom to look for and determine the appropriate treatment strategy. Notably, jaundice was found to be a significant predictor of malignancy by several studies [15,27,38,42]. In particular, a nomogram developed by Attiyeh et al. [15] automatically assigned a predicted probability of highrisk disease of "1" to patients with jaundice. Another symptom that showed high association with malignancy was weight loss. Among five studies that examined weight loss separately [15,19,20,33,42], all studies except one [33] found weight loss to be significantly associated with malignancy. While many symptoms depend on the patient's report and tend to be subjective,

wherein surgery is decided based on the EUS findings in patients with worrisome features.

The findings are in accordance with most of the widely used guidelines. The AGA guideline utilizes size, dilated main pancreatic duct, solid component, and positive cytology to determine the treatment strategy [6]. Reference size ≥ 3 cm, dilated main pancreatic duct, and associated solid component were considered risk factors, and presence of at least two of these would warrant endoscopic ultrasound-fine needle aspiration (EUS-FNA). In the case of positive cytology or presence of a solid component and a dilated pancreatic duct, surgery is indicated. Since this guideline is for asymptomatic neoplastic pancreatic cysts, the symptoms were not considered.

Unlike the conservative AGA guidelines, the European study group proposes a more aggressive approach in BD-IPMN patients [7,8]. Presence of jaundice, positive cytology, enhancing mural nodule (≥5 mm), solid mass, and main pancreatic duct ≥10 mm are absolute indicators for surgery. Growth rate ≥5 mm/year, elevated serum CA 19-9 level, main pancreatic duct dilatation between 5–9.9 mm, cyst diameter ≥40 mm, new onset diabetes mellitus, acute pancreatitis, and enhancing mural nodule (<5 mm) are relative indicators wherein healthy patients may opt for surgery.

The IAP guidelines stratify the features into high-risk stigmata and worrisome features. The high-risk stigmata and worrisome features warrant surgery and EUS, respectively. High-risk stigmata include obstructive jaundice in a patient with cystic lesion of the head of the pancreas, enhancing mural nodule ≥5 mm, and main pancreatic duct ≥10 mm. Worrisome features include cyst ≥3 cm, enhancing mural nodule <5 mm, thickened/enhancing cyst walls, main duct size 5–9 mm, abrupt change in caliber of pancreatic duct with distal pancreatic atrophy, lymphadenopathy, increased CA 19-9 serum level, and cystic growth rate ≥5 mm/2 years. The features used by the IAP and European study group are similar. However, the IAP guidelines are slightly more conservative, wherein surgery is decided based on the EUS findings in patients with worrisome features.

In this study, all the features were included to validate those featuring in various guidelines. In addition, other features such as locularity, multiplicity, and CEA serum level were explored. The parameter of symptoms showed a significant association with malignancy. However, the symptoms could be heterogenous and often vague. They consisted of one or combinations of clinical findings such as abdominal pain, weight loss, pancreatitis, and jaundice. Therefore, it is difficult to define what symptom to look for and determine the appropriate treatment strategy. Notably, jaundice was found to be a significant predictor of malignancy by several studies [15,27,38,42]. In particular, a nomogram developed by Attiyeh et al. [15] automatically assigned a predicted probability of high-risk disease of "1" to patients with jaundice. Another symptom that showed high association with malignancy was weight loss. Among five studies that examined weight loss separately [15,19,20,33,42], all studies except one [33] found weight loss to be significantly associated with malignancy. While many symptoms depend on the patient's report and tend to be subjective, jaundice and weight loss are symptoms that can be objectively quantified. Therefore, instead of considering symptoms as a whole, utilizing jaundice and weight loss to predict malignancy seemed reasonable, and studies defining the cut-off values for these symptoms should be warranted. Nevertheless, jaundice and weight loss are symptoms often associated with overt cancer and may have limited value in predicting earlier malignant transformation such as high-grade dysplasia. Our results showed that 27.4% of asymptomatic patients reported malignancy, demonstrating that absence of symptoms does not assure the absence of malignancy. Therefore, radiologic and biochemical changes may be more important in early detection of malignant transformations.

Previously, a cyst size of 3 cm was considered an absolute indication of BD-IPMN [10,51,52]. However, subsequent studies found that size alone was insufficient to predict malignancy, and although size correlated with malignancy risk, the safe cut-off limit was unclear [53–56]. The European study group does not consider a cyst size of 3 cm as an absolute indication, but rather considers the presence of other risk factors as determining factors, unless the diameter reaches 4 cm [7]. The IAP also stepped down the 3-cm size criteria from an absolute indication to a worrisome feature since the 2012 consensus guidelines [1,9]. Likewise, although the AGA states that size ≥3 cm increases the risk of malignancy by three times [57], size is not the sole determinant of the strategy [6]. In the present study, size ≥3 cm

increased the malignancy risk by two times. Although size is a significant factor, its impact is not as great as that of other features. Hence, size alone has a limited potential in predicting malignancy.

Wall thickening is a feature considered exclusively in the guidelines by IAP. It was introduced in the 2012 consensus guideline [1]. In this study, wall thickening increased the malignancy risk by 2.5 times. However, it is uncertain whether the wall thickening was accompanied by enhancement in the studies. Other cystic characteristics such as multilocularity or multiplicity of cysts did not increase the risk of malignancy.

Mural nodule is one of the strongest and most consistent risk factors in all the guidelines. Mural nodule is an absolute indication according to the European study group, and it could be an indication if it is accompanied by main pancreatic duct dilatation >5 mm according to the AGA guidelines. The AGA found that solid component increased the risk by almost eight times after reviewing 816 patients in seven studies [57]. They found that the incidence of malignancy was 73% in patients with a mural nodule as compared to 23% in those without a mural nodule. In 25 studies with 4495 patients, the malignancy rate was 52.5% in those with a mural nodule and 20.4% in those without. Furthermore, this study found that the malignancy risk was four times higher in patients with mural nodule. Nevertheless, mural nodule is one of the highly predictive factors of malignancy. Recently, enhancement and size of mural nodule have received attention, and these factors were applied to the IAP and European study group guidelines [1,8,58–61]. However, the diagnostic performances vary according to the imaging modality used, and meta-analysis cannot be conducted with the limited number of studies. Therefore, this study did not sub-analyze the mural nodule feature by size or enhancement, and future studies are needed to clarify the effect of these factors.

The main pancreatic duct change is another consistent risk factor of malignancy. The European study group, AGA, and IAP guidelines include main pancreatic duct dilatation. The AGA did not provide a definition of main duct dilation, whereas the European study group and IAP defined duct dilatation as dilatation >5 mm [1,6–9,57]. The IAP and European study group further stratified the risk level according to the extent of dilatation. Main duct dilation between 5–9 mm and >1 cm is considered as worrisome feature and high-risk stigmata by the IAP, or as a relative and absolute indication by the European study group. In contrast, the AGA requires that the solid component be accompanied by main duct dilatation for it to qualify as an indication. Interestingly, the AGA did not find a significant association between dilated pancreatic duct and malignancy (OR, 2.38, 95% CI 0.71–8.00), but included it in their guidelines because the review was performed with resected IPMNs [6,57]. Nevertheless, the main duct dilatation is a well-recognized risk factor that was also confirmed in this meta-analysis. However, the reference cut-off values vary according to studies, and each guideline weighs the same criteria differently. Future efforts are required to reach a consensus. Another change often studied and considered in the IAP guidelines is the abrupt change in caliber. Although this may overlap with main duct dilatation and may be considered an extreme form of dilatation, its OR was the second highest in this study at 7.41. However, this was based only on four studies, and the true predictive value needs further validation.

Lymphadenopathy was recently added to the IAP guidelines during the 2017 revision [9]. There are no references to lymphadenopathy in the European study group or AGA guidelines. Although least attention was given to lymphadenopathy, it demonstrated the highest OR, showing 8.5 times increased risk of malignancy. There were only four studies with a pooled cohort of 390 patients, of which 6.2% had lymphadenopathy. More studies are needed to accurately evaluate the impact of lymphadenopathy in predicting the malignancy in BD-IPMN cases.

Finally, a biochemical marker, CA 19-9, was indicated as a relative risk factor in the European guidelines and as a worrisome feature in the revised 2017 IAP guidelines [7,9]. This study showed that elevated CA 19-9 above 37 U/mL had four times higher risk of malignancy, which is similar to the risk associated with mural nodule. In addition to CA 19-9, the role of CEA was examined, which posed twice the risk of malignancy when elevated above 5 ng/mL. However, only four studies were examined and its actual role needs to be further studied for a definitive conclusion.

There are several limitations in this study. First, all the studies included in this meta-analysis were observational studies, and potential biases are likely to be greater in such studies. Thus, the results should always be interpreted with caution. Second, the studies were conducted on resected IPMNs, thus limiting the knowledge regarding the natural course of the disease. Conversely, this ensures the most accurate pathologic diagnosis. Third, some features had slightly different or more specified definitions in the guidelines, e.g., for "enhanced" wall thickening, and different values for duct dilatations. However, for analysis with an adequate population, the features could not be too narrowly defined. Finally, as the risk of malignancy is likely to increase, an analysis of risk by combination of features or creating a predictive model would have been informative.

#### **4. Materials and Methods**

#### *4.1. Literature Search Strategy*

A literature search was conducted using the MEDLINE to identify a relevant study about the outcomes in patients with worrisome features or high-risk stigmata of IPMN and malignancy proven by surgery or biopsy. A combination of search terms, including IPMN, computed tomography (CT), magnetic resonance image (MRI), EUS, malignancy, worrisome features, or predictive features, were used.

#### *4.2. Inclusion*/*Exclusion Criteria*

Studies were included if they met the following criteria: written in English, full-article, publication year between January 2000 and May 2019, patient with BD-IPMN diagnosed by CT, MRI or EUS and final pathological diagnosis by surgical resection or biopsy, and >10 patients in the study. We excluded case reports, case series with small sample size (<10 patients), review articles, editorials, consensus proceedings, studies without pathological diagnosis, not within field of interest, and insufficient or overlapping data.

#### *4.3. Data Extraction and Quality Assessment*

Two reviewers (W.K. and Y.H.) independently extracted the data from each study and resolved their disagreements by discussion or by consulting a third reviewer (J-Y.J.). The following data were collected from the studies that met the criteria. (1) Study—publication year, study design, and study location. (2) Cases—total number of BD-IPMN patients, frequency of pathologic malignancy in BD-IPMN, age, and sex. (3) Cystic morphology—maximum cyst size, presence of mural nodules, and maximum diameter of main pancreatic duct. (4) Clinical data—symptoms (jaundice, diabetes, pain, and weight loss), imaging methods, CA 19-9 level (normal value 0–37 U/mL), and CEA level (normal value 0–5 ng/mL). (5) Outcomes—cytology result and pathology result.

Malignant BD-IPMN was identified when there was histological evidence of BD-IPMN with invasive carcinoma or high-grade dysplasia after surgical resection, and cytological/histological evidence of high-grade dysplasia/malignant cells was found after FNA/biopsy of BD-IPMN with or without associated radiological signs of malignancy.

The choice of the articles included in this review were in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement (PRISMA) [62], and a PRISMA flowchart was formulated (Figure 1) for transparency of the conclusions reached by the authors. The quality of included studies was assessed using the Newcastle Ottawa Scale [63] by two reviewers (W.K. and Y.H.).

#### *4.4. Data Analysis*

Interpretative analysis of the OR between positive and negative worrisome features in IPMN patients was performed. The OR of BD-IPMN with or without worrisome features/high-risk stigmata was calculated by dividing the total number of events by the total number of patients. If these specific data were not provided in a study, it was calculated by adding or subtracting the number of patients who had confirmed pathology and imaging data. The corresponding 95% CIs were calculated using exact methods. A meta-analysis of all eligible studies identified was then planned with the Review Manager software (RevMan) (version 5.3; The Cochrane Collaboration, The Nordic Cochrane Center, Copenhagen, Denmark) using a random-effects model. This model was used because we believe that the relevant variation in the risk is most likely a consequence of inter-study differences. Statistical analysis was performed for all stages of this meta-analysis in accordance with the MOOSE guidelines [64]. The quantity of heterogeneity and publication bias was assessed. A *p*-value < 0.050 was accepted as statistically significant.

#### **5. Conclusions**

This study examined the parameters used to predict malignancy as specified by the most commonly used guidelines. This not only included clinical and radiographic features, but also biochemical features. The results confirmed that all the currently used features are valid. However, each guideline utilizes certain features and weighs the impact of each feature differently, resulting in different treatment strategies in BD-IPMN patients presenting similar features. This study hopes to contribute in making future guidelines more compatible and standardized.

**Author Contributions:** Conceptualization, W.K. and J.-Y.J.; methodology, Y.H., Y.B., J.S.K., Y.J.C.; software, Y.H.; validation, Y.H., W.K. and J.-Y.J.; formal analysis, Y.H. and W.K.; investigation, W.K. and J.-Y.J.; resources, H.K. and J.-Y.J.; data curation, Y.B., J.S.K., Y.J.C., Y.H., and H.K.; writing—original draft preparation, W.K. and Y.H.; writing—review and editing, W.K., Y.H., H.K., and J.-Y.J.; visualization, W.K. and Y.H.; supervision, J.-Y.J.; project administration, Y.H.; funding acquisition, J.-Y.J. All authors have read and agreed to the published version of the manuscript.

**Funding:** This study was supported by grant no. 23-2017-0090 from the SNUH Research Fund.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

#### *Systematic Review*

## **Ductal Dilatation of** ≥**5 mm in Intraductal Papillary Mucinous Neoplasm Should Trigger the Consideration for Pancreatectomy: A Meta-Analysis and Systematic Review of Resected Cases**

**Y.H. Andrew Wu 1,†, Atsushi Oba 1,2,† , Laurel Beaty 1,3, Kathryn L. Colborn 1,3,4 , Salvador Rodriguez Franco 1,5 , Ben Harnke <sup>6</sup> , Cheryl Meguid <sup>1</sup> , Daniel Negrini 1,7 , Roberto Valente 1,8 , Steven Ahrendt 1,9 , Richard D. Schulick 1,9 and Marco Del Chiaro 1,9,\***


**Simple Summary:** Intraductal papillary mucinous neoplasms (IPMN) are common but difficult to manage since accurate tools for diagnosing malignancy are unavailable. This study evaluates the diagnostic value of main pancreatic duct (MPD) diameter for detecting IPMN malignancy, using a meta-analysis of published data. The result suggests that malignancy is highly prevalent in IPMN with ductal dilatation of >5 mm.

**Abstract:** Intraductal papillary mucinous neoplasms (IPMN) are common but difficult to manage since accurate tools for diagnosing malignancy are unavailable. This study tests the diagnostic value of the main pancreatic duct (MPD) diameter for detecting IPMN malignancy using a meta-analysis of published data of resected IPMNs. Collected from a comprehensive literature search, the articles included in this analysis must report malignancy cases (high-grade dysplasia (HGD) and invasive carcinoma (IC)) and MPD diameter so that two MPD cut-offs could be created. The sensitivity, specificity, and odds ratios of the two cutoffs for predicting malignancy were calculated. A review of 1493 articles yielded 20 retrospective studies with 3982 resected cases. A cutoff of ≥5 mm is more sensitive than the ≥10 mm cutoff and has pooled sensitivity of 72.20% and 75.60% for classification of HGD and IC, respectively. Both MPD cutoffs of ≥5 mm and ≥10 mm were associated with malignancy (OR = 4.36 (95% CI: 2.82, 6.75) vs. OR = 3.18 (95% CI: 2.25, 4.49), respectively). The odds of HGD and IC for patients with MPD ≥5 mm were 5.66 (95% CI: 3.02, 10.62) and 7.40 (95% CI: 4.95, 11.06), respectively. OR of HGD and IC for MPD ≥10 mm cutoff were 4.36 (95% CI: 3.20, 5.93) and

**Citation:** Wu, Y.H.A.; Oba, A.; Beaty, L.; Colborn, K.L.; Rodriguez Franco, S.; Harnke, B.; Meguid, C.; Negrini, D.; Valente, R.; Ahrendt, S.; et al. Ductal Dilatation of ≥5 mm in Intraductal Papillary Mucinous Neoplasm Should Trigger the Consideration for Pancreatectomy: A Meta-Analysis and Systematic Review of Resected Cases. *Cancers* **2021**, *13*, 2031. https://doi.org/ 10.3390/cancers13092031

Academic Editors: Niccola Funel and Sohei Satoi

Received: 2 March 2021 Accepted: 19 April 2021 Published: 22 April 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

4.75 (95% CI: 2.39, 9.45), respectively. IPMN with MPD of >5 mm could very likely be malignant. In selected IPMN patients, pancreatectomy should be considered when MPD is >5 mm.

**Keywords:** pancreatic main duct dilatation; intraductal papillary mucinous neoplasm; high grade dysplasia; invasive carcinoma; pancreatic cystic neoplasm; pancreatic cancer; meta-analysis

#### **1. Introduction**

Pancreatic cystic lesions are common. The prevalence of these lesions is around 50% in the general population and increases with age [1]. Of these cystic lesions, pancreatic cystic neoplasm (PCN) management is challenging and important in modern pancreatology. The challenge lies in the difficulty in accurately discerning completely benign PCNs from ones with potential for malignancy progression [2–5].

Intraductal papillary mucinous neoplasm (IPMN) is a pre-cancerous lesion that accounts for at least half of all PCNs [6]. Morphologically, IPMNs can be divided into two major categories: (1) the branch-duct IPMN (BD-IPMN) that only involves the peripheral pancreatic ducts, and (2) main-duct IPMN (MD-IPMN) and mixed-type IPMN that involve the main pancreatic duct (MPD) and/or the branch ducts [7]. The morphologic and radiologic classification for identifying BD-IPMN, MD-IPMN, or mixed-type IPMN is extremely important. It is well known that BD-IPMNs have a very low risk for cancer progression; in contrast, MD- and mixed-type IPMNs are more prone to becoming cancerous [2,8]. Studies have shown that it is safe to surveil small BD-IPMNs (below 3–4 cm) that do not have any radiological features suggestive of malignancy [8–10]. In contrast, IPMNs involving the MPDs are more aggressive and generally need to be treated surgically [6,11].

The International Consensus Guidelines and the European Guidelines for managing IPMNs concordantly suggest an aggressive approach for managing MD- and mixed-type IPMNs. However, the role of MPD dilatation in deciding the indication for surgical resection of IPMN is moderately different in the two guidelines. The International Consensus Guidelines for IPMN management suggests that MD- or mixed-type IPMN with MPD dilatation of 10 mm or more should be referred for surgical treatment (high-risk stigmata) [11]. However, the same guideline suggests that surgery could still be considered for patients with 5–9.9 mm MPD (a worrisome feature), if there is the presence of mural nodule(s) ≥5 mm, cytology positivity for malignancy, or main-duct features suspicious for malignancy involvement [11]. The first European Guideline, published in 2013, recommended lowering the cutoff of MPD dilatation for surgery indication from 10 mm to 6 mm [2]. This approach was later supported by retrospective analysis that showed the implementation of MPD dilatation of 10 mm as a cutoff for surgery indication risks the possibility of IPMN undertreatment, as patients could have already developed invasive carcinoma (IC) or high-grade dysplasia (HGD) [12–14]. For better management of IPMNs, the latest evidence-based European Guidelines, published in 2018, recommends surgical resection of IPMN with MPD dilatation of 5–9.9 mm, if the patient is fit and has a long life expectancy (relative indications for surgery) [6]. This approach has also been recently supported by large retrospective surgical analyses [15].

Meanwhile, some studies have demonstrated a safe conservative approach in managing patients with suspected MD- or mixed-type IPMNs that have MPD dilatation of 5–9.9 mm [16,17]. Despite the compelling data, those studies need to be reassessed since high-volume centers have demonstrated that the accuracy of the pre-operative diagnosis of PCNs is approximately 60–80% [18–20]. Therefore, a high percentage of the suspected MD- or mixed-type IPMNs in those studies could very likely represent other more benign pancreatic diseases, such as chronic pancreatitis [18–20]. Hence, the optimal study design to analyze the disease status of PCNs is to evaluate a cohort with available histology data [21].

The role of MPD dilatation is important for establishing surgical indication. However, currently there is no consensus across existing guidelines regarding the degree of ductal

dilatation that warrants pancreatectomy. This study is designed to evaluate the association of HGD and IC with ≥5 mm and ≥10 mm MPD. Sensitivity and specificity for detecting HGD and IC were calculated using the two MPD cutoffs.

#### **2. Materials and Methods**

#### *2.1. Article Search/Selection and Outcome Assessment*

This systematic review and meta-analysis was performed by following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Guideline [22]. Literature search was performed on 28 February 2020 by a professional librarian at the Strauss Health Sciences Library. The following databases were queried: Ovid MEDLINE(R) ALL 1946 to 27 February 2020; Embase; Web of Science; Google Scholar. The search strategy focused on obtaining all existing literature related to MPD dilatation in IPMN and its association to malignancy. The terms including but not limited to: intraductal papillary mucinous or intraductal mucinous papillary or ipmn or ipmt; dilat or size or diameter or cut off or cutoff or mm or millimeter or milli meter were used for the systematic search. Full search strategy is provided in online Figure S1. The articles obtained from the search were uploaded to the Endnote version X9 citation management application, in which duplicated search results were eliminated. The final list of articles for initial review was uploaded to Covidence, a software designed for reviewing and selecting articles for systematic review and meta-analysis.

In the initial phase of reviewing articles, Y.H.A.W. and A.O. independently studied all records uploaded to Covidence. If the title or abstract of the articles were relevant to the topic of this study, the independent reviewers (Y.H.A.W. and A.O.) would thoroughly perform a screening of the entire article to assess for eligibility of inclusion. Non-English articles were excluded at the title/abstract review phase. Should there be a disagreement in eligibility of the included articles, D.N. would resolve the discrepancy.

Studies evaluating the association between MPD, IPMN (including BD-IPMN, MD-IPMN, or mixed-type IPMN), and malignancy were included in this study. We specifically only included studies that reported MPD ranges that could be used to create two cut-offs: ≥5 mm and ≥10 mm. Studies also had to include the counts or rates of malignancy and non-malignancy for each MPD category. We also required that included articles show histological diagnosis of lesions, specifically whether they were non-malignant or malignant. In this study, histology classification and definition are in accordance with the 2015 Baltimore Consensus Meeting Guideline [23] or the current World Health Organization guidelines [24] for IPMN. In short, HGD or IC were defined as malignancy, whereas lowgrade dysplasia or moderate-grade dysplasia were defined as non-malignancy. In the event of encountering two or more publications that had identical cohorts, only the publication with the most complete dataset was included in this study.

The primary outcomes of this study were the proportion of patients with HGD, IC, or malignancy.

#### *2.2. Data Extraction*

Study characteristics and data were independently extracted by two investigators (Y.H.A.W. and A.O.) and recorded on a standardized data extraction form. Any discrepancies were resolved by other reviewers (S.F., D.N., L.B., and K.L.C.). Data extracted and used for analysis are listed in Table 1. The final extracted data were reviewed by Y.H.A.W., A.O., S.F., L.B., and K.L.C.

#### *2.3. Risk of Bias Analysis*

Y.H.A.W., S.F., and A.O. assessed the quality of the included articles by conducting an evaluation using the Risk of Bias in Non-randomized Studies of Interventions (ROBINS-I) [41]. ROBINS-1 is designed to assess an article by evaluating 7 domains of risk of bias (Table S1). The assessment will grade the 7 domains of an article with low, moderate, serious, critical, or not assessable risk. In the end, the 7 domains were collectively analyzed

for every article included in this study. Additionally, Funnel plots and Egger's regression tests were used to assess publication bias (Figure S2).


**Table 1.** Characteristics of included articles.

\* Percentage of cases within MPD Dilatation category; Abbreviations: M = Malignancy; NM = Non-Malignancy.

#### *2.4. Data Synthesis and Statistical Analysis*

Tests of association: Random effects models were used to estimate adjusted diagnostic odds ratios (OR) and 95% confidence intervals for the pooled data. These were estimated using the meta package in RStudio (RStudio, Boston, Massachusetts), which provides functions for diagnostic meta-analysis [42]. A continuity correction was applied to all cells in a 2 × 2 table when necessary. We evaluated the association between cutoffs of ≥5 mm and ≥10 mm and classification of HGD, IC, and malignancy. There were three primary comparisons made for the two cutoffs: (1) classification of non-malignancy and malignancy; (2) classification of non-malignancy and HGD, and (3) classification of non-malignancy and IC. Therefore, we estimated six ORs using two cutoffs and three disease classifications. Forest Plots were used to compare individual-study ORs and the pooled OR. Forest plots include OR estimates from both the fixed effects model and the random effects model. Heterogeneity amongst the included publications was assessed using Cochrane's Q test and I<sup>2</sup> . Based on the results of these tests, this paper utilized only the random effects estimates for interpretations.

Diagnostic tests: Pooled sensitivities and specificities and 95% confidence intervals for the same six comparisons were also estimated using the mada package (R Foundation for Statistical Computing, Vienna, Austria). Using a random effects model, the mada package implements a bivariate estimation of sensitivity and specificity described by Reitsma et al. [43]. This bivariate approach is necessary because the sensitivity and specificity of a test are interrelated; therefore, univariate approaches to estimation are inappropriate. From these pooled values for sensitivity and specificity, we then estimated AUC for each comparison. The summary receiver operating characteristic (SROC) curves were plotted to assess the spread of diagnostic measures for each comparison and cutoff.

We conducted an additional analysis to evaluate the impact of including the large study by Del Chiaro et al. (senior author of this study) [15]. This was done by repeating the analyses described above excluding that study.

#### **3. Results**

The search strategy identified 3338 citations. After removing duplicated articles, 1493 were eligible for title and abstract review. Initial title and abstract review performed by two independent reviewers (Y.H.A.W. and A.O.) yielded 120 articles eligible for full-article review. Ultimately, 20 manuscripts were included for qualitative and quantitative analysis. The authors excluded one hundred manuscripts for the following reasons: 41 articles did not have MPD dilatation values; 37 articles did not have all the MPD ranges that adhered to our criteria; 8 articles had incorrect study design and did not include information needed for quantitative meta-analysis; 5 articles had vague or no histology diagnosis; 5 articles did not provide sufficient information for case number retrieval; 2 articles only used ultrasound/endoscopic ultrasound for preoperative evaluation; 1 article lacked surgical pathology results; 1 article had duplication of cohort with another included article (Figure 1).

**Figure 1.** PRISMA flow chart showing the article selection process.

Characteristics of articles included for qualitative and quantitative analysis are listed in Tables 1 and 2. From the 20 included articles [12,14,15,17,25–31,33–40,44,45], a total of 3982 resected IPMN cases (including BD-IPMN, MD-IPMN, or mixed-type IPMN) were collected. Of the 3982 resected cases, 1516 and 2466 were malignant and non-malignant cases, respectively. Of the 1516 malignant cases, 316, 680, and 520 cases had MPD range of <5 mm, 5–9.9 mm, and ≥10 mm, respectively. Similarly, of the 2466 non-malignant cases, 1404, 773, and 289 cases had MPD range of <5 mm, 5–9.9 mm, and ≥10 mm, respectively. Ten of the 20 studies included pathologically confirmed HGD and IC cases. Of the 818 malignant cases collected from the 10 articles, 421 and 397 cases had histology diagnosis of HGD and IC, respectively. Of the 421 HGD cases, 89, 187, and 145 cases had MPD range of <5 mm, 5–9 mm, and ≥10 mm, respectively. Similarly, of the 397 IC cases, 70, 182, 145 cases had MPD range of <5 mm, 5–9.9 mm, and ≥10 mm, respectively.

**Table 2.** Characteristics of included article with postoperative histology diagnosis of HGD and IC.


\* Percentage of cases within MPD Dilatation category; Abbreviations: HGD, High Grade Dysplasia; IC, Invasive Carcinoma; NM, Non-Malignancy.

> ROBINS-I indicated that the risk of bias of the included study was moderate in 18 [12,14,15,17,26–40] studies and serious in 2 [25,32]. The results of this analysis are provided in Table S1.

#### *3.1. Tests of Association*

Both the ≥5 mm and ≥10 mm cutoffs in resected cases were significantly associated with higher risk of malignancy compared to <5 mm and <10 mm, respectively (≥5 mm: OR = 4.36 [95% CI: 2.82, 6.75, 6.75, I<sup>2</sup> = 81.7%, Cochran's Q *<sup>p</sup>* < 0.0001]; <sup>≥</sup>10 mm: OR = 3.18 [95% CI: 2.25, 4.49, I<sup>2</sup> = 68.2%, Cochran's Q *p* < 0.0001]). The odds of HGD were over five times higher for patients with <sup>≥</sup>5 mm MPD {5.66 (95% CI: 3.02, 10.62, I<sup>2</sup> = 71.2%, Cochran's Q *p =* 0.002)} compared to patients with MPD < 5 mm and over four times higher for patients with <sup>≥</sup>10 mm MPD {4.36 (95% CI: 3.20, 5.93, I<sup>2</sup> = 8.4%, Cochran's Q *p =* 0.365)} compared to patients with <10 mm. The odds of IC were over seven times higher for patients with MPD <sup>≥</sup>5 mm {7.40 (95% CI: 4.95, 11.06, I<sup>2</sup> = 27.5%, Cochran's Q *p =* 0.2189)} compared to <5 mm and 4.7 times higher for patients with ≥10 mm duct {4.75 (95% CI: 2.39, 9.45, I<sup>2</sup> = 72.3%, Cochran's Q *p =* 0.0002)} compared to <10 mm. Forest Plots are shown in Figure 2. Summary of the odds of HGD and IC for different MPD cutoffs are in Table S2.

#### *3.2. Diagnostic Tests*

For all six comparisons, the tests for equality of sensitivities and specificities were all found to be significant (*p* < 0.001). This indicated heterogeneity between studies, and therefore we estimated pooled sensitivities and specificities using random effects models (Table 3). The six SROC curves from the random effects models are included in Figure 3.

**Figure 2.** Meta-analysis forest plot presented in Diagnostic Odds Ratio (OR): (**A**) malignancy (M) and non-malignancy (NM), cutoff of 5 mm; (**B**) malignancy (M) and non-malignancy (NM), cutoff of 10 mm; (**C**) high-grade dysplasia (HGD) and non-malignancy (NM), cutoff of 5 mm; (**D**) high-grade dysplasia (HGD) and non-malignancy (NM), cutoff of 10 mm; (**E**) invasive carcinoma (IC) vs. non-malignancy (NM), cutoff of 5 mm; (**F**) invasive carcinoma (IC) vs. non-malignancy (NM), cutoff of 10 mm.



Abbreviations: NM, Non-Malignancy.

**Figure 3.** Summary receiver operating characteristic (SROC) curves for the bivariate diagnostic tests. Dot is point estimate that represents estimated pooled sensitivity/1-specificity for the included studies. Circle represents the confidence interval around the point estimate.

Using a cutoff point of 5 mm, specificity was 58.6% and sensitivity was 74.8% for classification of malignancy. The AUC was 0.716. For the cutoff point of 10 mm, specificity was 86.4% and sensitivity were 33.8% for classification of malignancy. The AUC was 0.586. A cutoff of 5 mm had 70.1% specificity and 72.2% sensitivity for classification of HGD. The AUC was 0.769. A cutoff of 10 mm had 88.7% specificity and 35.7% sensitivity for classification of HGD. The AUC was 0.587. A cutoff point of 5 mm had 69.7% specificity and 75.6% sensitivity for classification of IC. The AUC was 0.786. A cutoff of 10 mm had 88.2% specificity and 36.6% sensitivity for classification of IC. The AUC was 0.587 (Figure 3).

#### *3.3. Subset Analysis Excluding Del Chiaro et al.*

The analysis excluding the large study by Del Chiaro et al. [15] yielded very similar results to the overall results presented in this study. The results of this analysis are provided in Table S2.

#### **4. Discussion**

The correct clinical management of IPMNs is crucial for the prevention of pancreatic cancer [46]. An overtreatment of low-grade dysplasia lesions could result in unnecessary morbidity and mortality related to pancreatic surgery. Surgery is indicated for IPMNs in an attempt to remove IC and HGD. The latter is the optimal pre-invasive histology form and time-point for surgical intervention. Unfortunately, there is no available method that can effectively discriminate HGD from IC, except a few experimental approaches that are not yet implemented in clinical practice [15,47,48]. Recent studies reported that pancreatectomy-related mortality has decreased from 7.3% since 2000 [49], and the benchmark for postoperative mortality after pancreaticoduodenectomy published in 2019 was ≤1.6% [50], which is drastically lower than the 17–42% probability of five-year survival of

resected IC [50–54]. Therefore, surgery performed on IC could be too late for extending the survival of patients, and it could be more beneficial to resect MD-IPMN while in HGD form.

According to the European evidence-based guideline for PCN management, IPMN MPD dilatation of 5–9.9 mm and ≥10 mm are relative and absolute indication for surgery, respectively [6]. Recent studies have demonstrated IPMN with MPD dilatation of ≥5 mm have a malignancy rate of 30–90% [12,15,28,33,55–59]. In this meta-analysis of 20 retrospective studies, malignancy was detected in nearly 46.8% of resected IPMN patients with 5–9.9 mm MPD and 63.5% of resected IPMN patients with ≥10 mm MPD. In total, 53.1% of resected IPMN patients with ≥5 mm MPD had malignancy. On the contrary, only 18.4% of resected cases with MPD < 5 mm were malignant IPMN. Pooled OR for malignancy calculated in this meta-analysis showed that MPD cutoff set at ≥5 mm was higher than ≥10 mm (OR = 4.4 vs. 3.2) in the resected cases. To improve the overall survival rate of malignant IPMN, HGD should be surgically removed before allowing it to progress to IC. Pooled OR from this meta-analysis showed that the odds of HGD were higher in ≥5 mm MPD than ≥10 mm MPD (OR = 5.7 vs. 4.4) in resected cases. Similarly, pooled OR for IC was also higher in ≥5 mm MPD than in ≥10 mm MPD (OR = 7.4 vs. 4.8). These data suggest that ductal dilatation of ≥5 mm should trigger the consideration for pancreatectomy.

The pooled sensitivities of ≥5 mm MPD in predicting IPMN HGD and IC were 72.2% and 75.6%, respectively. When using ≥5 mm MPD as cutoff, SROC AUC was 0.769 and 0.786 for HGD and IC, respectively, which were higher than those of ≥10 mm cutoff (AUC = 0.657 and 0. 587 for HGD and IC, respectively). The predictive role of ≥10 mm MPD for malignancy is unquestionable, but ≥5 mm MPD cutoff should also be considered as a highly sensitive factor for detecting HGD and/or IC. This lower cutoff could potentially identify malignancy in advance and improve survival of IPMN patients. However, Marchegiani et al. suggested that utilization of MPD dilatation as the sole indicator for pancreatectomy could encourage unnecessary surgical procedure and that most individuals with 5–9 mm MPD should be managed expectantly, performing conversion surgery when tumor progression is identified during close, regular follow-ups [17]. In the observation arm of the same study, 3 out of 46 (6.5%) IPMN patients with 5–9 mm MPD eventually underwent surgery [17]. However, Salvia et al. reported that the diagnostic accuracy for MD-IPMNs is only approximately 80% [19], implying some of those IPMN patients who underwent conservative treatment did not actually have IPMN [21]. If a PCN were to present with mural nodule(s) or cyst(s), fine-needle aspiration (FNA) with real-time endoscopic ultrasound (EUS) could be utilized to obtain histology proof and to make a diagnosis. It is important to note that EUS can accurately identify morphologic features of PCNs but is only 51% accurate in discerning the difference between mucinous and non-mucinous lesions [60]. The accuracy would increase up to 79% if CEA measurements were obtained from cystic fluid using EUS-FNA [60]. However, mural nodules and cysts ≥40 mm have only been reported in 1.2% to 21.3% and 23.2% to 27.5% of malignant IPMNs, respectively [13,15], and FNA would not have been a possible approach if ductal dilatation were the only cross-sectional imagining finding. Studies that include resected cases disproportionately favor inclusion of more serious cases that require surgery, but at least the diagnosis could be confirmed with surgical pathology. Including only pathologically confirmed IPMN cases for analysis, this study shows that many HGD and IC patients would be missed if the decision to operate were made at MPD ≥ 10 mm associated with low sensitivity.

The results in this study underline the need for a surgical evaluation for MPD ≥5 mm and are not conclusive evidence that suggest the risk for developing cancer is higher in IPMN with 5–9.9 mm MPD. On the other hand, it could also be possible that IPMN with MPD dilatation over 10 mm are slow progressive diseases that would turn invasive when MPD reaches larger dilatation. Using ≥5 mm MPD to select surgical candidates might also identify low-grade dysplasia. Considering that the highest incidence of IC occurs with MD-IPMN, we could assume that the surgical treatment of MD-IPMN with low-grade dysplasia is perhaps not an overtreatment when compared to surgical treatment of BD-IPMN with

low-grade dysplasia [21]. Nevertheless, it is important to know that the currently available information for IPMN management is not perfect, and it is important to diagnose IPMN with existing sensitive diagnostic tools to increase the overall survival rates of the patients, as the mean frequency of malignancy in MPD dilatation of ≥5 mm was 61.6% and that of IC IPMN was 43.1% [11].

Recognizing the potential for the Del Chiaro et al. study [15] of heavily influencing the results presented in this study given its overall size, we completed a full analysis of the data excluding this study. The results and conclusions were unchanged.

Our study has limitations related to the use of retrospective studies that lack conservatively managed cohorts as a control group. Cases could not be stratified based on the imaging modalities used. The most accurate cutoff line for MPD dilatation could not be determined in this study since data for MPD dilatation are presented in ranges. In addition, other features associated with IPMN malignancy, such as mural nodule, cytology, elevated CA 19–9, or cyst diameter, could not be included in this analysis [11]. In addition, this study includes only resected IPMNs. Therefore, the results of this study are not representative of the unresected IPMN population. However, this study was performed based on recently published study that showed MPD dilatation is currently the best predictor of HGD or IC in IPMN [15]. Although estimates of heterogeneity above 50% were observed for some of our analyses, we found no reason for publication bias based on visual inspection of the funnel plot and the results of the Egger's test (Egger Regression Model). Additionally, and as stated in the methodology, all of our statistical models were estimated using random model effects, which have previously been used successfully to account for this limitation [42]. Even with some limitations related to the retrospective nature of this study, this analysis contains the largest volume of IPMN cases analyzed and our methodological approach properly accounted for study-level variation.

#### **5. Conclusions**

MPD dilatation is an important predictive factor of IPMN malignancy and 5 mm is a highly sensitive cutoff that detects high-risk pre-cancerous or cancerous lesions in resected cases. It is important to note that this study cannot draw a conclusion for non-surgical cases. However, the need for pancreatectomy should be thoroughly evaluated in patients with ductal dilatation of ≥5 mm. The result of this study implies that MPD dilatation over 5 mm should trigger the referral of a patient to a high-volume center for further consultation. The decision to perform resection should only be considered after careful evaluation of multiple aspects related to the general conditions and the expectancy of life of a patient. In conjunction with new biomarkers or diagnostic modalities such as pancreatoscopy [61,62], MPD dilatation detection could improve surgical patient selection and reduce overall IPMN malignancy mortality.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/10 .3390/cancers13092031/s1. Figure S1: Detailed Search Strategy for Existing Online Literature, Figure S2: Funnel plot with Diagnostic Odds Ratio (DOR) and Egger's Test for publication bias assessment, Table S1: ROBINS-I Risk of Bias Assessment of included Articles, Table S2: Odds ratio, Sensitive, and Specificity for different comparisons and cutoffs.

**Author Contributions:** Y.H.A.W.; A.O.; L.B., K.L.C.; and M.D.C. were involved in the study design. Y.H.A.W.; A.O., L.B.; K.L.C., and B.H. were involved in management of this study. Y.H.A.W.; A.O.; L.B.; K.L.C.; B.H.; S.F. and D.N. performed data collection, analysis, and interpretation. Y.H.A.W.; A.O.; K.L.C.; and M.D.C. were involved in the initial drafting of this manuscript. C.M.; R.V.; S.A.; S.F. and R.D.S. revised and provided important intellectual content. The authors showed no conflict in the content of this manuscript upon submission. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Conflicts of Interest:** None of the authors have any personal conflict of interest to declare, except for Del Chiaro who is a co-principal investigator on a Boston Scientific study investigating the role of intraoperative pancreatoscopy in IPMN patients. Del Chiaro has also been awarded an industry grant from Haemonetics Inc. to conduct a multicenter study that evaluates the prognostic implications and changes of thromboelastographic (TEG) in pancreas cancer patients.

#### **References**


MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Cancers* Editorial Office E-mail: cancers@mdpi.com www.mdpi.com/journal/cancers

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18

www.mdpi.com ISBN 978-3-0365-1865-7